lstm_2.ipynb

{
 "nbformat": 4,
 "nbformat_minor": 0,
 "metadata": {
  "colab": {
   "provenance": [],
   "toc_visible": true
  },
  "kernelspec": {
   "name": "python3",
   "display_name": "Python 3"
  },
  "language_info": {
   "name": "python"
  }
 },
 "cells": [
  {
   "cell_type": "markdown",
   "source": [
    "# LSTM "
   ],
   "metadata": {
    "id": "kqgz1G8Io0La",
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "Resources: \n",
    "\n",
    "http://colah.github.io/posts/2015-08-Understanding-LSTMs/\n",
    "\n",
    "https://medium.com/analytics-vidhya/beginners-guide-to-stock-prediction-using-lstm-7010bf8b8c21\n",
    "\n",
    "https://towardsdatascience.com/deep-learning-model-interpretation-using-shap-a21786e91d16\n",
    "\n",
    "https://machinelearningmastery.com/reshape-input-data-long-short-term-memory-networks-keras/\n",
    "\n",
    "https://github.com/MohammadFneish7/Keras_LSTM_Diagram\n"
   ],
   "metadata": {
    "id": "1J_FgZVhpJZd",
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Imports"
   ],
   "metadata": {
    "id": "YI35qYb-o667",
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "interim"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "source": [
    "import os\n",
    "import time\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import pickle\n",
    "import shapely\n",
    "import random\n",
    "\n",
    "# import cartopy #Was throwing an error, couldn't tell you why?\n",
    "# import cartopy.crs as ccrs\n",
    "# import cartopy.feature as cfeature\n",
    "# from cartopy.mpl.gridliner import LONGITUDE_FORMATTER, LATITUDE_FORMATTER\n",
    "# import cartopy.io.img_tiles as cimgt\n",
    "import matplotlib.pyplot as plt\n",
    "# import cartopy.mpl.geoaxes\n",
    "\n",
    "import mpl_scatter_density # adds projection='scatter_density'\n",
    "from mpl_toolkits.axes_grid1.inset_locator import inset_axes\n",
    "import matplotlib as mpl\n",
    "from mpl_toolkits.axes_grid1 import make_axes_locatable\n",
    "from matplotlib.colors import ListedColormap, LinearSegmentedColormap\n",
    "from scipy.io import loadmat\n",
    "\n",
    "import sklearn.metrics as skm\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "from sklearn.metrics import mean_squared_error\n",
    "\n",
    "from keras.preprocessing.sequence import TimeseriesGenerator\n",
    "\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras"
   ],
   "metadata": {
    "id": "aNlgxfPxpLvc",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": 39,
   "outputs": []
  },
  {
   "cell_type": "code",
   "execution_count": 143,
   "outputs": [],
   "source": [
    "#Set up all lat/lon pairs that we will be transforming into arrays:\n",
    "def latlon_setup(df, n_input):\n",
    "    pixel_counts = df.groupby(['lat', 'lon']).size() #Counts\n",
    "    pixel_counts = pixel_counts.reset_index()\n",
    "    #Remove lat/lon sets that don't have enough records, as defined by our n_input variable\n",
    "    latlon = pd.lreshape(pixel_counts[pixel_counts[0] > (n_input -1)][['lat', 'lon']], {'lat':['lat'], 'lon':['lon']})\n",
    "    latlon = latlon.drop_duplicates()\n",
    "    #Make sure that we aren't sampling in a pattern:\n",
    "    latlon = latlon.sample(frac = 1, random_state = 0)\n",
    "    print(\"Length of df of unique latlon paris that meet cutoff: \", len(latlon)) #if 60: 123873, if 80: 85603\n",
    "    return latlon"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 144,
   "outputs": [],
   "source": [
    "def rand_gen_np(df, n_input):\n",
    "  \"\"\"Returns random windows of input data at 1/10th rate \"\"\"\n",
    "  all_windows = list(range(len(df) - n_input + 1)) #List of total number of records in the dataframe minus the n_input window size\n",
    "  n_samples = int(((len(df) - n_input + 1)/10) + 1) #all windows large enough not to be weeded out get at least one sample\n",
    "  # make slices of data\n",
    "  Xs, ys = [], []\n",
    "\n",
    "  for w in range(n_samples): #For every sample we want to take:\n",
    "      window = random.choice(all_windows) #Randomly select which window to take\n",
    "      all_windows.remove(window) #Remove selection from list to prevent resampling\n",
    "      current_slice = df.iloc[window:window+n_input]\n",
    "      X = current_slice[['snowtemp', 'soiltemp', 'omega', 'Tbh', 'Tbv', 'cf', 'h']].values\n",
    "      y = current_slice[['density', 'voc']].values[-1]\n",
    "      Xs.append(X)\n",
    "      ys.append(y)\n",
    "\n",
    "  X = np.stack(Xs)\n",
    "  y = np.stack(ys)\n",
    "  return X, y"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "outputs": [],
   "source": [
    "#What about using only numpy to pre-process?\n",
    "# https://stackoverflow.com/questions/66353849/confusions-about-the-input-shape-in-lstm-models\n",
    "\n",
    "def gen_np(df, n_input):\n",
    "  \"\"\"Produces sequential windows of n_input historical records of data from the provided dataframe.\"\"\"\n",
    "  num_windows = len(df) - n_input + 1\n",
    "  # make slices of data\n",
    "  Xs, ys = [], []\n",
    "\n",
    "  for w in range(num_windows):\n",
    "      current_slice = df.iloc[w:w+n_input]\n",
    "      X = current_slice[['snowtemp', 'soiltemp', 'omega', 'Tbh', 'Tbv', 'cf', 'h']].values\n",
    "      y = current_slice[['density', 'voc']].values[-1]\n",
    "      Xs.append(X)\n",
    "      ys.append(y)\n",
    "\n",
    "  X = np.stack(Xs)\n",
    "  y = np.stack(ys)\n",
    "  return X, y"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 152,
   "outputs": [],
   "source": [
    "def data_prep(bigdf, latlon, path, n_input, n_features, X, y, counter = 0, savept = 0):\n",
    "    start = time.strftime(\"%Y-%m-%d %H:%M:%S\", time.localtime())\n",
    "    print(\"Started at: \", start)\n",
    "    for i in range(len(latlon)-counter): #Iterate though our set of viable lat/lons, minus counter to restart if needed.\n",
    "      counter += 1\n",
    "      df = bigdf[(bigdf['lat'] == latlon['lat'].iloc[i]) & (bigdf['lon'] == latlon['lon'].iloc[i])] #Make a sub df of each lat/lon pair\n",
    "      df = df.sort_values(by = 'date') #Make sure we have our values correctly ordered\n",
    "      X2, y2 = rand_gen_np(df, n_input) #Make your arrays, in this case random\n",
    "      #print(\"Dimensions: \", X2.shape, y2.shape)\n",
    "      X = np.vstack([X, X2]) #Append your arrays\n",
    "      y = np.concatenate([y, y2])\n",
    "      if counter%10000 == 0: #For every ten thousand lat/long pairs acts as check point\n",
    "        print(\"Counter: \", counter, \", Shape X: \", X.shape, time.strftime(\"%Y-%m-%d %H:%M:%S\", time.localtime()))\n",
    "        with open((path + str(savept) + '.npy'), 'wb') as f: #Saves two arrays in same file, arrays can be opened separately.\n",
    "          np.save(f, X)\n",
    "          np.save(f, y)\n",
    "        print(\"Wipeout: \", savept)\n",
    "        savept += 1\n",
    "        X = np.empty([0, n_input, n_features])\n",
    "        y = np.empty([0,2])\n",
    "    print(\"Save last: \")\n",
    "    with open((path + '_last.npy'), 'wb') as f: #final save of X,y to grab any remaining lat/lon pairs since last checkpoint:\n",
    "          np.save(f, X)\n",
    "          np.save(f, y)\n",
    "    print(\"Done! And it only took from \", start, \" to \", time.strftime(\"%Y-%m-%d %H:%M:%S\", time.localtime()))"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "outputs": [],
   "source": [
    "#X_test = np.empty([0, n_input, n_features])\n",
    "#y_test = np.empty([0,2])\n",
    "#path = 'LSTM_Data_60Sparse_Test'\n",
    "def assemble_test(X_test, y_test, path):\n",
    "    for filename in os.listdir(path):\n",
    "        p = os.path.join(path, filename)\n",
    "        with open(p, 'rb') as f:\n",
    "            X_load = np.load(f)\n",
    "            y_load = np.load(f)\n",
    "        X_test = np.vstack([X_test, X_load]) #Append\n",
    "        y_test = np.concatenate([y_test, y_load])\n",
    "        print(\"Test: \", f, X_test.shape)\n",
    "    return X_test, y_test"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 154,
   "outputs": [],
   "source": [
    "#X_train = np.empty([0, n_input, n_features])\n",
    "#y_train = np.empty([0,2])\n",
    "#X_val = np.empty([0, n_input, n_features])\n",
    "#y_val = np.empty([0,2])\n",
    "#path = 'LSTMData80Sparse'\n",
    "#switch = int(len(os.listdir(path))*.2)\n",
    "def assemble_train(X_train, y_train, X_val, y_val, path, switch):\n",
    "    filecount = len(os.listdir(path))\n",
    "    for filename in os.listdir(path):\n",
    "        filecount = filecount-1\n",
    "        f = os.path.join(path, filename)\n",
    "        if filecount > switch:\n",
    "            print(f)\n",
    "            with open(f, 'rb') as f: #Saves two arrays in same file, arrays can be opened separately or together.\n",
    "                X_load = np.load(f)\n",
    "                y_load = np.load(f)\n",
    "            X_train = np.vstack([X_train, X_load]) #Append your arrays\n",
    "            y_train = np.concatenate([y_train, y_load])\n",
    "            print(\"Train: \", f)\n",
    "        else:\n",
    "            with open(f, 'rb') as f: #Saves two arrays in same file, arrays can be opened separately or together.\n",
    "                X_load = np.load(f)\n",
    "                y_load = np.load(f)\n",
    "            X_val = np.vstack([X_val, X_load]) #Append your arrays\n",
    "            y_val = np.concatenate([y_val, y_load])\n",
    "            print(\"Val: \", f)\n",
    "    return X_train, y_train, X_val, y_val"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "outputs": [],
   "source": [
    "#https://towardsdatascience.com/choosing-the-right-hyperparameters-for-a-simple-lstm-using-keras-f8e9ed76f046\n",
    "\n",
    "#Adam has a default of learning_rate = 0.001, beta_1 = 0.9, beta_2 = 0.999, amsgrad = False.\n",
    "\n",
    "def model_lstm():\n",
    "  model = keras.models.Sequential()\n",
    "  #1st layer:\n",
    "  model.add(keras.layers.LSTM(60, activation='relu', return_sequences=True, input_shape=(n_input, n_features)))\n",
    "  model.add(keras.layers.Dropout(0.2))\n",
    "  #2nd layer:\n",
    "  model.add(keras.layers.LSTM(60))\n",
    "  model.add(keras.layers.Dropout(0.2))\n",
    "  #Output layer:\n",
    "  model.add(keras.layers.Dense(2))\n",
    "  model.compile(optimizer='adam', loss='mean_squared_error')\n",
    "  #model.summary()\n",
    "  return(model)\n",
    "\n"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "outputs": [],
   "source": [
    "def get_metrics(y_test, y_pred):\n",
    "    mae = skm.mean_absolute_error(y_test[:,0], y_pred[:,0])\n",
    "    mse = skm.mean_squared_error(y_test[:,0], y_pred[:,0])\n",
    "    rmse = skm.mean_squared_error(y_test[:,0], y_pred[:,0], squared = False)\n",
    "    r2 = skm.r2_score(y_test[:,0], y_pred[:,0])\n",
    "    print(\"Snow Density Performance: \")\n",
    "    print(\"MAE: \", mae)\n",
    "    print(\"MSE: \", mse)\n",
    "    print(\"RMSE: \", rmse)\n",
    "    print(\"R2: \", r2)\n",
    "\n",
    "    mae2 = skm.mean_absolute_error(y_test[:, 1], y_pred[:, 1])\n",
    "    mse2 = skm.mean_squared_error(y_test[:, 1], y_pred[:, 1])\n",
    "    rmse2 = skm.mean_squared_error(y_test[:, 1], y_pred[:, 1], squared=False)\n",
    "    r22 = skm.r2_score(y_test[:, 1], y_pred[:, 1])\n",
    "    print(\"VOD Performance: \")\n",
    "    print(\"MAE: \", mae2)\n",
    "    print(\"MSE: \", mse2)\n",
    "    print(\"RMSE: \", rmse2)\n",
    "    print(\"R2: \", r22)\n",
    "\n",
    "    return mae, mse, rmse, r2, mae2, mse2, rmse2, r22"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Data Prep Example"
   ],
   "metadata": {
    "id": "91eqxcM0zBC7",
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "source": [
    "# #DataPrep: Scaling\n",
    "# data = pd.read_parquet('/content/drive/Shareddrives/AI For Earth/Datasets/processed_data_with_datetime.parquet')\n",
    "# scaler = MinMaxScaler() \n",
    "# data_scaled = data[['density', 'snowtemp', 'soiltemp', 'omega', 'Tbh', 'Tbv', 'cf', 'h', 'voc']] #We want lat, lon and date unscaled, as these will be used for sorting, not training. \n",
    "# data_scaled = pd.DataFrame(scaler.fit_transform(data_scaled), columns=data_scaled.columns)\n",
    "# data_scaled['lat'] = data['lat']\n",
    "# data_scaled['lon'] = data['lon']\n",
    "# data_scaled['date'] = data['date']\n",
    "# display(data_scaled.head())\n",
    "# data_scaled.to_parquet('/content/drive/Shareddrives/AI For Earth/Datasets/scaled_processed_data_LatLonDateUnscaled.parquet')\n",
    "# del(data)\n",
    "# # Test/Train split:\n",
    "# train_scaled = data_scaled[data_scaled['date']<'2020-01-01']\n",
    "# display(train_scaled.tail())\n",
    "# train_scaled.to_parquet('/content/drive/Shareddrives/AI For Earth/Datasets/train_scaled_LatLonDateUnscaled.parquet')\n",
    "# del(train_scaled)\n",
    "# test_scaled = data_scaled[data_scaled['date']>='2020-01-01']\n",
    "# display(test_scaled.head())\n",
    "# test_scaled.to_parquet('/content/drive/Shareddrives/AI For Earth/Datasets/test_scaled_LatLonDateUnscaled.parquet')\n",
    "# del(test_scaled)\n",
    "# del(data_scaled)"
   ],
   "metadata": {
    "id": "xWxrdir83NIN",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": []
  },
  {
   "cell_type": "code",
   "source": [
    "# Import data:\n",
    "train = pd.read_parquet('train_scaled_LatLonDateUnscaled.parquet')\n",
    "\n",
    "print(train.columns)\n",
    "\n",
    "print(train['date'].max())\n",
    "print(train['date'].min())\n",
    "print(len(train))\n",
    "print(train.info())\n",
    "train.head()"
   ],
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 671
    },
    "id": "UaBAYRGFy_dK",
    "outputId": "b134e610-0353-4a46-b455-5a639e004146",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Index(['density', 'snowtemp', 'soiltemp', 'omega', 'Tbh', 'Tbv', 'cf', 'h',\n",
      "       'voc', 'lat', 'lon', 'date'],\n",
      "      dtype='object')\n",
      "2019-12-31 00:00:00\n",
      "2019-01-01 00:00:00\n",
      "18329587\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 18329587 entries, 0 to 18329586\n",
      "Data columns (total 12 columns):\n",
      " #   Column    Dtype         \n",
      "---  ------    -----         \n",
      " 0   density   float64       \n",
      " 1   snowtemp  float64       \n",
      " 2   soiltemp  float64       \n",
      " 3   omega     float64       \n",
      " 4   Tbh       float64       \n",
      " 5   Tbv       float64       \n",
      " 6   cf        float64       \n",
      " 7   h         float64       \n",
      " 8   voc       float64       \n",
      " 9   lat       float64       \n",
      " 10  lon       float64       \n",
      " 11  date      datetime64[ns]\n",
      "dtypes: datetime64[ns](1), float64(11)\n",
      "memory usage: 1.8 GB\n",
      "None\n"
     ]
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "    density  snowtemp  soiltemp     omega       Tbh       Tbv        cf  \\\n",
       "0  0.166630  0.908916  0.998302  0.779338  0.710750  0.708752  0.305672   \n",
       "1  0.176869  0.955726  0.995804  0.783470  0.715650  0.708749  0.311526   \n",
       "2  0.180525  0.959748  0.996218  0.790908  0.722035  0.714708  0.322348   \n",
       "3  0.183057  0.966050  0.996963  0.790081  0.721543  0.716810  0.302173   \n",
       "4  0.167732  0.926476  0.995828  0.789999  0.709735  0.705815  0.303099   \n",
       "\n",
       "          h       voc        lat        lon       date  \n",
       "0  0.124832  0.240164  62.684110 -160.44087 2019-01-01  \n",
       "1  0.123756  0.228918  62.532246 -160.44087 2019-01-01  \n",
       "2  0.123569  0.242246  62.381150 -160.44087 2019-01-01  \n",
       "3  0.123513  0.240669  62.230810 -160.44087 2019-01-01  \n",
       "4  0.123518  0.239886  62.684110 -160.34750 2019-01-01  "
      ],
      "text/html": [
       "\n",
       "  <div id=\"df-878a72aa-8a2b-4688-b434-2d1210f116f4\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>density</th>\n",
       "      <th>snowtemp</th>\n",
       "      <th>soiltemp</th>\n",
       "      <th>omega</th>\n",
       "      <th>Tbh</th>\n",
       "      <th>Tbv</th>\n",
       "      <th>cf</th>\n",
       "      <th>h</th>\n",
       "      <th>voc</th>\n",
       "      <th>lat</th>\n",
       "      <th>lon</th>\n",
       "      <th>date</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.166630</td>\n",
       "      <td>0.908916</td>\n",
       "      <td>0.998302</td>\n",
       "      <td>0.779338</td>\n",
       "      <td>0.710750</td>\n",
       "      <td>0.708752</td>\n",
       "      <td>0.305672</td>\n",
       "      <td>0.124832</td>\n",
       "      <td>0.240164</td>\n",
       "      <td>62.684110</td>\n",
       "      <td>-160.44087</td>\n",
       "      <td>2019-01-01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.176869</td>\n",
       "      <td>0.955726</td>\n",
       "      <td>0.995804</td>\n",
       "      <td>0.783470</td>\n",
       "      <td>0.715650</td>\n",
       "      <td>0.708749</td>\n",
       "      <td>0.311526</td>\n",
       "      <td>0.123756</td>\n",
       "      <td>0.228918</td>\n",
       "      <td>62.532246</td>\n",
       "      <td>-160.44087</td>\n",
       "      <td>2019-01-01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.180525</td>\n",
       "      <td>0.959748</td>\n",
       "      <td>0.996218</td>\n",
       "      <td>0.790908</td>\n",
       "      <td>0.722035</td>\n",
       "      <td>0.714708</td>\n",
       "      <td>0.322348</td>\n",
       "      <td>0.123569</td>\n",
       "      <td>0.242246</td>\n",
       "      <td>62.381150</td>\n",
       "      <td>-160.44087</td>\n",
       "      <td>2019-01-01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.183057</td>\n",
       "      <td>0.966050</td>\n",
       "      <td>0.996963</td>\n",
       "      <td>0.790081</td>\n",
       "      <td>0.721543</td>\n",
       "      <td>0.716810</td>\n",
       "      <td>0.302173</td>\n",
       "      <td>0.123513</td>\n",
       "      <td>0.240669</td>\n",
       "      <td>62.230810</td>\n",
       "      <td>-160.44087</td>\n",
       "      <td>2019-01-01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.167732</td>\n",
       "      <td>0.926476</td>\n",
       "      <td>0.995828</td>\n",
       "      <td>0.789999</td>\n",
       "      <td>0.709735</td>\n",
       "      <td>0.705815</td>\n",
       "      <td>0.303099</td>\n",
       "      <td>0.123518</td>\n",
       "      <td>0.239886</td>\n",
       "      <td>62.684110</td>\n",
       "      <td>-160.34750</td>\n",
       "      <td>2019-01-01</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-878a72aa-8a2b-4688-b434-2d1210f116f4')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-878a72aa-8a2b-4688-b434-2d1210f116f4 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-878a72aa-8a2b-4688-b434-2d1210f116f4');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ]
     },
     "metadata": {},
     "execution_count": 7
    }
   ]
  },
  {
   "cell_type": "markdown",
   "source": [
    "Whatever we pick as n_input (number of historical records) will be our cutoff value - any pixels that have fewer than n_input records can not be run. \n",
    "\n",
    "If we were to set n_input to three, we would go from 467957 pixels to 428025. A minimal loss. "
   ],
   "metadata": {
    "id": "AvzhHfb7_JiU",
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "Similar to this problem: \n",
    "https://stackoverflow.com/questions/61177311/creating-a-timeseriesgenerator-with-multiple-inputs"
   ],
   "metadata": {
    "id": "GtSFonlZKOIS",
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "source": [
    "n_input = 3 #how many samples/rows/timesteps to look in the past in order to forecast the next sample\n",
    "\n",
    "pixel_counts = train.groupby(['lat', 'lon']).size()\n",
    "pixel_counts = pixel_counts.reset_index()\n",
    "print(pixel_counts.head())\n",
    "\n",
    "#There are 467957 pixel locations in our set. Get all unique pairs of lat lon:\n",
    "\n",
    "print(pixel_counts[pixel_counts[0] > n_input])\n",
    "print(len(pixel_counts[pixel_counts[0] > n_input]))\n",
    "\n",
    "latlon = pd.lreshape(pixel_counts[pixel_counts[0] > n_input][['lat', 'lon']], {'lat':['lat'], 'lon':['lon']})\n",
    "\n",
    "latlon = latlon.drop_duplicates()\n",
    "print(len(latlon))\n",
    "latlon.head()"
   ],
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 756
    },
    "id": "clIQVSpF0UrT",
    "outputId": "b10d4466-22e1-4ca4-d12a-7db6fdc03aec",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "         lat        lon  0\n",
      "0 -54.977610 -67.920130  1\n",
      "1 -54.855858 -67.826770  3\n",
      "2 -54.855858 -67.733406  2\n",
      "3 -54.855858 -67.546684  2\n",
      "4 -54.855858 -67.453316  1\n"
     ]
    },
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "             lat        lon  0\n",
       "10     -54.73447 -68.106850  5\n",
       "11     -54.73447 -68.013490  5\n",
       "12     -54.73447 -67.826770  5\n",
       "13     -54.73447 -67.733406  4\n",
       "15     -54.73447 -67.546684  4\n",
       "...          ...        ... ..\n",
       "467952  81.24827 -19.559128  5\n",
       "467953  81.24827 -19.465767  5\n",
       "467954  81.24827 -19.372406  5\n",
       "467955  81.24827 -19.279045  5\n",
       "467956  81.24827 -19.185684  5\n",
       "\n",
       "[413624 rows x 3 columns]"
      ],
      "text/html": [
       "\n",
       "  <div id=\"df-5fd8ccd3-6fe1-4b12-b0ff-8423492e2b1c\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>lat</th>\n",
       "      <th>lon</th>\n",
       "      <th>0</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-68.106850</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-68.013490</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-67.826770</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-67.733406</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-67.546684</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>467952</th>\n",
       "      <td>81.24827</td>\n",
       "      <td>-19.559128</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>467953</th>\n",
       "      <td>81.24827</td>\n",
       "      <td>-19.465767</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>467954</th>\n",
       "      <td>81.24827</td>\n",
       "      <td>-19.372406</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>467955</th>\n",
       "      <td>81.24827</td>\n",
       "      <td>-19.279045</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>467956</th>\n",
       "      <td>81.24827</td>\n",
       "      <td>-19.185684</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>413624 rows × 3 columns</p>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-5fd8ccd3-6fe1-4b12-b0ff-8423492e2b1c')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-5fd8ccd3-6fe1-4b12-b0ff-8423492e2b1c button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-5fd8ccd3-6fe1-4b12-b0ff-8423492e2b1c');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ]
     },
     "metadata": {}
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "413624\n",
      "413624\n"
     ]
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "        lat        lon\n",
       "0 -54.73447 -68.106850\n",
       "1 -54.73447 -68.013490\n",
       "2 -54.73447 -67.826770\n",
       "3 -54.73447 -67.733406\n",
       "4 -54.73447 -67.546684"
      ],
      "text/html": [
       "\n",
       "  <div id=\"df-7efbd2c1-b1c4-455b-b563-04d8079b950c\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>lat</th>\n",
       "      <th>lon</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-68.106850</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-68.013490</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-67.826770</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-67.733406</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-67.546684</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-7efbd2c1-b1c4-455b-b563-04d8079b950c')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-7efbd2c1-b1c4-455b-b563-04d8079b950c button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-7efbd2c1-b1c4-455b-b563-04d8079b950c');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ]
     },
     "metadata": {},
     "execution_count": 31
    }
   ]
  },
  {
   "cell_type": "code",
   "source": [
    "#Try getting all values for a specific lat/lon pair:\n",
    "train1 = train[(train['lat'] == latlon['lat'].iloc[0]) & (train['lon'] == latlon['lon'].iloc[0])]\n",
    "print(len(train1))\n",
    "print(train1['date'].min())\n",
    "print(train1['date'].max())\n",
    "print(train1.head())\n",
    "#Looks good. \n",
    "\n",
    "#Get all values for a second lat/lon pair:\n",
    "train2 = train[(train['lat'] == latlon['lat'].iloc[1]) & (train['lon'] == latlon['lon'].iloc[1])]\n",
    "print(len(train2))\n",
    "print(train2['date'].min())\n",
    "print(train2['date'].max())\n",
    "print(train2.head())"
   ],
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 503
    },
    "id": "UKcO8Ak5_MNp",
    "outputId": "67177fc3-fbb1-4a88-f33e-037bc7b87585",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "5\n",
      "2019-08-11 00:00:00\n",
      "2019-08-19 00:00:00\n"
     ]
    },
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "           density  snowtemp  soiltemp  omega       Tbh       Tbv        cf  \\\n",
       "13562138  0.412166  0.816258  0.998811    0.7  0.668379  0.677059  0.319354   \n",
       "13563663  0.417339  0.805035  0.996234    0.7  0.678995  0.684643  0.319354   \n",
       "13565561  0.428953  0.870483  0.999304    0.7  0.677228  0.677272  0.319354   \n",
       "13566803  0.436613  0.892357  0.993725    0.7  0.677911  0.682764  0.319354   \n",
       "13568994  0.461971  0.903925  0.998780    0.7  0.672902  0.674774  0.319354   \n",
       "\n",
       "                 h       voc       lat       lon       date  \n",
       "13562138  0.133275  0.703391 -54.73447 -68.10685 2019-08-11  \n",
       "13563663  0.133275  0.703552 -54.73447 -68.10685 2019-08-12  \n",
       "13565561  0.133275  0.703702 -54.73447 -68.10685 2019-08-14  \n",
       "13566803  0.133275  0.703682 -54.73447 -68.10685 2019-08-15  \n",
       "13568994  0.133275  0.703603 -54.73447 -68.10685 2019-08-19  "
      ],
      "text/html": [
       "\n",
       "  <div id=\"df-013f9447-7fad-467a-93ff-a101c27030d4\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>density</th>\n",
       "      <th>snowtemp</th>\n",
       "      <th>soiltemp</th>\n",
       "      <th>omega</th>\n",
       "      <th>Tbh</th>\n",
       "      <th>Tbv</th>\n",
       "      <th>cf</th>\n",
       "      <th>h</th>\n",
       "      <th>voc</th>\n",
       "      <th>lat</th>\n",
       "      <th>lon</th>\n",
       "      <th>date</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>13562138</th>\n",
       "      <td>0.412166</td>\n",
       "      <td>0.816258</td>\n",
       "      <td>0.998811</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.668379</td>\n",
       "      <td>0.677059</td>\n",
       "      <td>0.319354</td>\n",
       "      <td>0.133275</td>\n",
       "      <td>0.703391</td>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-68.10685</td>\n",
       "      <td>2019-08-11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13563663</th>\n",
       "      <td>0.417339</td>\n",
       "      <td>0.805035</td>\n",
       "      <td>0.996234</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.678995</td>\n",
       "      <td>0.684643</td>\n",
       "      <td>0.319354</td>\n",
       "      <td>0.133275</td>\n",
       "      <td>0.703552</td>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-68.10685</td>\n",
       "      <td>2019-08-12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13565561</th>\n",
       "      <td>0.428953</td>\n",
       "      <td>0.870483</td>\n",
       "      <td>0.999304</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.677228</td>\n",
       "      <td>0.677272</td>\n",
       "      <td>0.319354</td>\n",
       "      <td>0.133275</td>\n",
       "      <td>0.703702</td>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-68.10685</td>\n",
       "      <td>2019-08-14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13566803</th>\n",
       "      <td>0.436613</td>\n",
       "      <td>0.892357</td>\n",
       "      <td>0.993725</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.677911</td>\n",
       "      <td>0.682764</td>\n",
       "      <td>0.319354</td>\n",
       "      <td>0.133275</td>\n",
       "      <td>0.703682</td>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-68.10685</td>\n",
       "      <td>2019-08-15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13568994</th>\n",
       "      <td>0.461971</td>\n",
       "      <td>0.903925</td>\n",
       "      <td>0.998780</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.672902</td>\n",
       "      <td>0.674774</td>\n",
       "      <td>0.319354</td>\n",
       "      <td>0.133275</td>\n",
       "      <td>0.703603</td>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-68.10685</td>\n",
       "      <td>2019-08-19</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-013f9447-7fad-467a-93ff-a101c27030d4')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-013f9447-7fad-467a-93ff-a101c27030d4 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-013f9447-7fad-467a-93ff-a101c27030d4');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ]
     },
     "metadata": {}
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "5\n",
      "2019-08-11 00:00:00\n",
      "2019-08-19 00:00:00\n"
     ]
    },
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "           density  snowtemp  soiltemp  omega       Tbh       Tbv        cf  \\\n",
       "13562142  0.412067  0.817112  0.995011    0.7  0.668841  0.675598  0.321291   \n",
       "13563678  0.416563  0.807155  0.992390    0.7  0.677074  0.683424  0.321291   \n",
       "13565563  0.426910  0.870855  0.998475    0.7  0.673722  0.678564  0.321291   \n",
       "13566807  0.433939  0.892694  0.990807    0.7  0.677054  0.680754  0.321291   \n",
       "13569000  0.457130  0.910077  0.997865    0.7  0.671337  0.676375  0.321291   \n",
       "\n",
       "                 h       voc       lat       lon       date  \n",
       "13562142  0.130967  0.703391 -54.73447 -68.01349 2019-08-11  \n",
       "13563678  0.130967  0.703552 -54.73447 -68.01349 2019-08-12  \n",
       "13565563  0.130967  0.703702 -54.73447 -68.01349 2019-08-14  \n",
       "13566807  0.130967  0.703682 -54.73447 -68.01349 2019-08-15  \n",
       "13569000  0.130967  0.703603 -54.73447 -68.01349 2019-08-19  "
      ],
      "text/html": [
       "\n",
       "  <div id=\"df-d27473d2-0cef-46fd-9a8e-90a94bc3ff24\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>density</th>\n",
       "      <th>snowtemp</th>\n",
       "      <th>soiltemp</th>\n",
       "      <th>omega</th>\n",
       "      <th>Tbh</th>\n",
       "      <th>Tbv</th>\n",
       "      <th>cf</th>\n",
       "      <th>h</th>\n",
       "      <th>voc</th>\n",
       "      <th>lat</th>\n",
       "      <th>lon</th>\n",
       "      <th>date</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>13562142</th>\n",
       "      <td>0.412067</td>\n",
       "      <td>0.817112</td>\n",
       "      <td>0.995011</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.668841</td>\n",
       "      <td>0.675598</td>\n",
       "      <td>0.321291</td>\n",
       "      <td>0.130967</td>\n",
       "      <td>0.703391</td>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-68.01349</td>\n",
       "      <td>2019-08-11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13563678</th>\n",
       "      <td>0.416563</td>\n",
       "      <td>0.807155</td>\n",
       "      <td>0.992390</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.677074</td>\n",
       "      <td>0.683424</td>\n",
       "      <td>0.321291</td>\n",
       "      <td>0.130967</td>\n",
       "      <td>0.703552</td>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-68.01349</td>\n",
       "      <td>2019-08-12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13565563</th>\n",
       "      <td>0.426910</td>\n",
       "      <td>0.870855</td>\n",
       "      <td>0.998475</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.673722</td>\n",
       "      <td>0.678564</td>\n",
       "      <td>0.321291</td>\n",
       "      <td>0.130967</td>\n",
       "      <td>0.703702</td>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-68.01349</td>\n",
       "      <td>2019-08-14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13566807</th>\n",
       "      <td>0.433939</td>\n",
       "      <td>0.892694</td>\n",
       "      <td>0.990807</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.677054</td>\n",
       "      <td>0.680754</td>\n",
       "      <td>0.321291</td>\n",
       "      <td>0.130967</td>\n",
       "      <td>0.703682</td>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-68.01349</td>\n",
       "      <td>2019-08-15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13569000</th>\n",
       "      <td>0.457130</td>\n",
       "      <td>0.910077</td>\n",
       "      <td>0.997865</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.671337</td>\n",
       "      <td>0.676375</td>\n",
       "      <td>0.321291</td>\n",
       "      <td>0.130967</td>\n",
       "      <td>0.703603</td>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-68.01349</td>\n",
       "      <td>2019-08-19</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-d27473d2-0cef-46fd-9a8e-90a94bc3ff24')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-d27473d2-0cef-46fd-9a8e-90a94bc3ff24 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-d27473d2-0cef-46fd-9a8e-90a94bc3ff24');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ]
     },
     "metadata": {}
    }
   ]
  },
  {
   "cell_type": "code",
   "source": [
    "# #To illistrate how Generators works include a reset index:\n",
    "# train1 = train1.reset_index(drop = True)\n",
    "# train1 = train1.reset_index()\n",
    "# print(train1.head())\n",
    "\n",
    "# train2 = train2.reset_index(drop = True)\n",
    "# train2 = train2.reset_index()\n",
    "# train2.head()"
   ],
   "metadata": {
    "id": "iuVBaE1jJvtd",
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 438
    },
    "outputId": "7e179f2a-4f2c-4434-d9ff-db404594dd93",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "   index   density  snowtemp  soiltemp     omega       Tbh       Tbv  \\\n",
      "0      0  0.166630  0.908916  0.998302  0.779338  0.710750  0.708752   \n",
      "1      1  0.203359  0.553834  0.998184  0.779338  0.714580  0.707774   \n",
      "2      2  0.191918  0.581473  0.992350  0.779338  0.713668  0.711221   \n",
      "3      3  0.194654  0.425676  0.988548  0.779338  0.711498  0.705567   \n",
      "4      4  0.196457  0.282219  0.980238  0.779338  0.714373  0.709466   \n",
      "\n",
      "         cf         h       voc       lat        lon       date  \n",
      "0  0.305672  0.124832  0.240164  62.68411 -160.44087 2019-01-01  \n",
      "1  0.305672  0.124832  0.240019  62.68411 -160.44087 2019-01-08  \n",
      "2  0.305672  0.124832  0.240003  62.68411 -160.44087 2019-01-09  \n",
      "3  0.305672  0.124832  0.239988  62.68411 -160.44087 2019-01-10  \n",
      "4  0.305672  0.124832  0.239974  62.68411 -160.44087 2019-01-11  \n"
     ]
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "   index   density  snowtemp  soiltemp    omega       Tbh       Tbv        cf  \\\n",
       "0      0  0.176869  0.955726  0.995804  0.78347  0.715650  0.708749  0.311526   \n",
       "1      1  0.200845  0.564406  0.994576  0.78347  0.718182  0.716408  0.311526   \n",
       "2      2  0.188242  0.584642  0.988450  0.78347  0.710517  0.714051  0.311526   \n",
       "3      3  0.190667  0.460968  0.984354  0.78347  0.714821  0.706018  0.311526   \n",
       "4      4  0.192554  0.298409  0.975456  0.78347  0.714592  0.709420  0.311526   \n",
       "\n",
       "          h       voc        lat        lon       date  \n",
       "0  0.123756  0.228918  62.532246 -160.44087 2019-01-01  \n",
       "1  0.123756  0.228975  62.532246 -160.44087 2019-01-08  \n",
       "2  0.123756  0.228984  62.532246 -160.44087 2019-01-09  \n",
       "3  0.123756  0.228993  62.532246 -160.44087 2019-01-10  \n",
       "4  0.123756  0.229001  62.532246 -160.44087 2019-01-11  "
      ],
      "text/html": [
       "\n",
       "  <div id=\"df-c7ccf670-04b5-4606-aa6b-78c805175f7d\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>index</th>\n",
       "      <th>density</th>\n",
       "      <th>snowtemp</th>\n",
       "      <th>soiltemp</th>\n",
       "      <th>omega</th>\n",
       "      <th>Tbh</th>\n",
       "      <th>Tbv</th>\n",
       "      <th>cf</th>\n",
       "      <th>h</th>\n",
       "      <th>voc</th>\n",
       "      <th>lat</th>\n",
       "      <th>lon</th>\n",
       "      <th>date</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>0.176869</td>\n",
       "      <td>0.955726</td>\n",
       "      <td>0.995804</td>\n",
       "      <td>0.78347</td>\n",
       "      <td>0.715650</td>\n",
       "      <td>0.708749</td>\n",
       "      <td>0.311526</td>\n",
       "      <td>0.123756</td>\n",
       "      <td>0.228918</td>\n",
       "      <td>62.532246</td>\n",
       "      <td>-160.44087</td>\n",
       "      <td>2019-01-01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>0.200845</td>\n",
       "      <td>0.564406</td>\n",
       "      <td>0.994576</td>\n",
       "      <td>0.78347</td>\n",
       "      <td>0.718182</td>\n",
       "      <td>0.716408</td>\n",
       "      <td>0.311526</td>\n",
       "      <td>0.123756</td>\n",
       "      <td>0.228975</td>\n",
       "      <td>62.532246</td>\n",
       "      <td>-160.44087</td>\n",
       "      <td>2019-01-08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>0.188242</td>\n",
       "      <td>0.584642</td>\n",
       "      <td>0.988450</td>\n",
       "      <td>0.78347</td>\n",
       "      <td>0.710517</td>\n",
       "      <td>0.714051</td>\n",
       "      <td>0.311526</td>\n",
       "      <td>0.123756</td>\n",
       "      <td>0.228984</td>\n",
       "      <td>62.532246</td>\n",
       "      <td>-160.44087</td>\n",
       "      <td>2019-01-09</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>0.190667</td>\n",
       "      <td>0.460968</td>\n",
       "      <td>0.984354</td>\n",
       "      <td>0.78347</td>\n",
       "      <td>0.714821</td>\n",
       "      <td>0.706018</td>\n",
       "      <td>0.311526</td>\n",
       "      <td>0.123756</td>\n",
       "      <td>0.228993</td>\n",
       "      <td>62.532246</td>\n",
       "      <td>-160.44087</td>\n",
       "      <td>2019-01-10</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>0.192554</td>\n",
       "      <td>0.298409</td>\n",
       "      <td>0.975456</td>\n",
       "      <td>0.78347</td>\n",
       "      <td>0.714592</td>\n",
       "      <td>0.709420</td>\n",
       "      <td>0.311526</td>\n",
       "      <td>0.123756</td>\n",
       "      <td>0.229001</td>\n",
       "      <td>62.532246</td>\n",
       "      <td>-160.44087</td>\n",
       "      <td>2019-01-11</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-c7ccf670-04b5-4606-aa6b-78c805175f7d')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-c7ccf670-04b5-4606-aa6b-78c805175f7d button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-c7ccf670-04b5-4606-aa6b-78c805175f7d');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ]
     },
     "metadata": {},
     "execution_count": 11
    }
   ]
  },
  {
   "cell_type": "code",
   "source": [
    "#Lets try reshaping as a tensor - LSTM takes in a 3d df:  number of samples (row count - n_input) by number of timesteps (rows) by the number of features (columns)\n",
    "# 41 time steps, 8 features (9 after removing y, lat, lon and date but adding index). sample number depends on n_input size (# of records used to make a prediction)\n",
    "\n",
    "x_train1 = train1[['index','snowtemp', 'soiltemp', 'omega', 'Tbh', 'Tbv', 'cf', 'h', 'voc']] #Swap comment if you are using index \n",
    "#x_train1 = train1[['snowtemp', 'soiltemp', 'omega', 'Tbh', 'Tbv', 'cf', 'h', 'voc']]\n",
    "x_train1 = np.array(x_train1)\n",
    "y_train1 = train1[['density']]\n",
    "y_train1 = np.array(y_train1)\n",
    "y_train1 = y_train1.reshape(-1) #now is values in one array, not an array of arrays size 1. \n",
    "\n",
    "print(\"set 1:\")\n",
    "print(x_train1.shape)\n",
    "print(y_train1.shape)\n",
    "\n",
    "#Second set\n",
    "x_train2 = train2[['index','snowtemp', 'soiltemp', 'omega', 'Tbh', 'Tbv', 'cf', 'h', 'voc']]\n",
    "#x_train2 = train2[['snowtemp', 'soiltemp', 'omega', 'Tbh', 'Tbv', 'cf', 'h', 'voc']]\n",
    "x_train2 = np.array(x_train2)\n",
    "y_train2 = train2[['density']]\n",
    "y_train2 = np.array(y_train2)\n",
    "y_train2 = y_train2.reshape(-1) #now is values in one array, not an array of arrays size 1. \n",
    "\n",
    "print(\"set 2:\")\n",
    "print(x_train2.shape)\n",
    "print(y_train2.shape)\n",
    "\n",
    "print(train1.head())\n",
    "print(y_train1[:5])"
   ],
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "pNxj2UqY_5So",
    "outputId": "0b040c04-a6f0-4876-8562-4dd272f1eff4",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "set 1:\n",
      "(41, 9)\n",
      "(41,)\n",
      "set 2:\n",
      "(46, 9)\n",
      "(46,)\n"
     ]
    }
   ]
  },
  {
   "cell_type": "markdown",
   "source": [
    "Important note on processing: We are trying to predict the *current* density from our input variables, so there is no need to move our y_data one forward. "
   ],
   "metadata": {
    "id": "vZkEzeA9d-gz",
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "source": [
    "#https://machinelearningmastery.com/how-to-use-the-timeseriesgenerator-for-time-series-forecasting-in-keras/\n",
    "\n",
    "n_features= x_train1.shape[1] # how many predictors/Xs/features we have to predict y\n",
    "b_size = 50 # Number of timeseries samples in each batch. It will self limit to feasable size, so you can go over and it will self correct with no error\n",
    "\n",
    "generator = TimeseriesGenerator(x_train1, y_train1, length=n_input, batch_size=b_size)\n",
    "generator2 = TimeseriesGenerator(x_train2, y_train2, length=n_input, batch_size=b_size)\n",
    "\n",
    "print(\"shape generator 1:\", generator[0][0].shape)\n",
    "print(\"shape generator 1:\", generator2[0][0].shape)\n",
    "\n",
    "print(\"generator 1 content:\")\n",
    "generator[0][0]"
   ],
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "kHsJ0QYXFLXm",
    "outputId": "501ca333-8baf-41ad-f6e8-7c1f0d567663",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "shape generator 1: (38, 3, 9)\n",
      "shape generator 1: (43, 3, 9)\n",
      "generator 1 content:\n"
     ]
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "array([[[ 0.        ,  0.90891553,  0.99830206, ...,  0.3056722 ,\n",
       "          0.12483246,  0.24016423],\n",
       "        [ 1.        ,  0.55383357,  0.99818396, ...,  0.3056722 ,\n",
       "          0.12483246,  0.24001892],\n",
       "        [ 2.        ,  0.58147313,  0.99235019, ...,  0.3056722 ,\n",
       "          0.12483246,  0.24000276]],\n",
       "\n",
       "       [[ 1.        ,  0.55383357,  0.99818396, ...,  0.3056722 ,\n",
       "          0.12483246,  0.24001892],\n",
       "        [ 2.        ,  0.58147313,  0.99235019, ...,  0.3056722 ,\n",
       "          0.12483246,  0.24000276],\n",
       "        [ 3.        ,  0.42567649,  0.98854799, ...,  0.3056722 ,\n",
       "          0.12483246,  0.23998776]],\n",
       "\n",
       "       [[ 2.        ,  0.58147313,  0.99235019, ...,  0.3056722 ,\n",
       "          0.12483246,  0.24000276],\n",
       "        [ 3.        ,  0.42567649,  0.98854799, ...,  0.3056722 ,\n",
       "          0.12483246,  0.23998776],\n",
       "        [ 4.        ,  0.28221916,  0.98023832, ...,  0.3056722 ,\n",
       "          0.12483246,  0.23997391]],\n",
       "\n",
       "       ...,\n",
       "\n",
       "       [[35.        ,  0.35802195,  0.96498329, ...,  0.3056722 ,\n",
       "          0.12483246,  0.24004565],\n",
       "        [36.        ,  0.31382154,  0.95995388, ...,  0.3056722 ,\n",
       "          0.12483246,  0.24004565],\n",
       "        [37.        ,  0.32477769,  0.9544777 , ...,  0.3056722 ,\n",
       "          0.12483246,  0.24004565]],\n",
       "\n",
       "       [[36.        ,  0.31382154,  0.95995388, ...,  0.3056722 ,\n",
       "          0.12483246,  0.24004565],\n",
       "        [37.        ,  0.32477769,  0.9544777 , ...,  0.3056722 ,\n",
       "          0.12483246,  0.24004565],\n",
       "        [38.        ,  0.41980394,  0.95392768, ...,  0.3056722 ,\n",
       "          0.12483246,  0.24004565]],\n",
       "\n",
       "       [[37.        ,  0.32477769,  0.9544777 , ...,  0.3056722 ,\n",
       "          0.12483246,  0.24004565],\n",
       "        [38.        ,  0.41980394,  0.95392768, ...,  0.3056722 ,\n",
       "          0.12483246,  0.24004565],\n",
       "        [39.        ,  0.49600264,  0.95516315, ...,  0.3056722 ,\n",
       "          0.12483246,  0.24004565]]])"
      ]
     },
     "metadata": {},
     "execution_count": 17
    }
   ]
  },
  {
   "cell_type": "code",
   "source": [
    "generator.targets #Generator targets is just a stright reproduction of y_train"
   ],
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "ptuzZ_raY-p1",
    "outputId": "99983098-f2ee-4163-d15e-b5282b0fc9ad",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "array([0.16662979, 0.20335938, 0.19191794, 0.19465384, 0.196457  ,\n",
       "       0.19744315, 0.18722819, 0.17017889, 0.20717552, 0.21762509,\n",
       "       0.24080997, 0.22371299, 0.22260159, 0.1603912 , 0.16048403,\n",
       "       0.16058131, 0.15255736, 0.10457073, 0.14727694, 0.15503449,\n",
       "       0.15932876, 0.17938672, 0.19660069, 0.17692231, 0.16101366,\n",
       "       0.15264193, 0.14955569, 0.12343908, 0.14789623, 0.18848583,\n",
       "       0.30205393, 0.28712824, 0.27823069, 0.27691329, 0.27262028,\n",
       "       0.2756086 , 0.27673971, 0.27772458, 0.27585466, 0.25279503,\n",
       "       0.24062114])"
      ]
     },
     "metadata": {},
     "execution_count": 15
    }
   ]
  },
  {
   "cell_type": "code",
   "source": [
    "#Lets try appending these generator objects:\n",
    "\n",
    "#We can use the generator object to build out our moving windows along each set,\n",
    "# then concatinate the arrays inside of it into one large dataset for training. \n",
    "gen = np.vstack([generator[0][0], generator2[0][0]])\n",
    "ygen_noprocess = np.concatenate([generator.targets, generator2.targets])\n",
    "ygen = np.concatenate([generator.targets[1:-(n_input-1)], generator2.targets[1:-(n_input-1)]])\n",
    "\n",
    "print(len(ygen_noprocess)) #This one is six longer, since there was an overhang of 3 extra values for each generator object. \n",
    "print(len(ygen))\n",
    "\n",
    "print(gen.shape)\n",
    "gen"
   ],
   "metadata": {
    "id": "F10JKAkqIMvY",
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "outputId": "f3664563-cf4d-4d54-e11c-564c21834ee9",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "87\n",
      "81\n",
      "(81, 3, 9)\n"
     ]
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "array([[[ 0.        ,  0.90891553,  0.99830206, ...,  0.3056722 ,\n",
       "          0.12483246,  0.24016423],\n",
       "        [ 1.        ,  0.55383357,  0.99818396, ...,  0.3056722 ,\n",
       "          0.12483246,  0.24001892],\n",
       "        [ 2.        ,  0.58147313,  0.99235019, ...,  0.3056722 ,\n",
       "          0.12483246,  0.24000276]],\n",
       "\n",
       "       [[ 1.        ,  0.55383357,  0.99818396, ...,  0.3056722 ,\n",
       "          0.12483246,  0.24001892],\n",
       "        [ 2.        ,  0.58147313,  0.99235019, ...,  0.3056722 ,\n",
       "          0.12483246,  0.24000276],\n",
       "        [ 3.        ,  0.42567649,  0.98854799, ...,  0.3056722 ,\n",
       "          0.12483246,  0.23998776]],\n",
       "\n",
       "       [[ 2.        ,  0.58147313,  0.99235019, ...,  0.3056722 ,\n",
       "          0.12483246,  0.24000276],\n",
       "        [ 3.        ,  0.42567649,  0.98854799, ...,  0.3056722 ,\n",
       "          0.12483246,  0.23998776],\n",
       "        [ 4.        ,  0.28221916,  0.98023832, ...,  0.3056722 ,\n",
       "          0.12483246,  0.23997391]],\n",
       "\n",
       "       ...,\n",
       "\n",
       "       [[40.        ,  0.41487661,  0.95611865, ...,  0.31152562,\n",
       "          0.12375563,  0.22951882],\n",
       "        [41.        ,  0.37513617,  0.95061769, ...,  0.31152562,\n",
       "          0.12375563,  0.22951882],\n",
       "        [42.        ,  0.45427506,  0.94472445, ...,  0.31152562,\n",
       "          0.12375563,  0.22951882]],\n",
       "\n",
       "       [[41.        ,  0.37513617,  0.95061769, ...,  0.31152562,\n",
       "          0.12375563,  0.22951882],\n",
       "        [42.        ,  0.45427506,  0.94472445, ...,  0.31152562,\n",
       "          0.12375563,  0.22951882],\n",
       "        [43.        ,  0.51633741,  0.94439164, ...,  0.31152562,\n",
       "          0.12375563,  0.22951882]],\n",
       "\n",
       "       [[42.        ,  0.45427506,  0.94472445, ...,  0.31152562,\n",
       "          0.12375563,  0.22951882],\n",
       "        [43.        ,  0.51633741,  0.94439164, ...,  0.31152562,\n",
       "          0.12375563,  0.22951882],\n",
       "        [44.        ,  0.5841771 ,  0.94703269, ...,  0.31152562,\n",
       "          0.12375563,  0.22951882]]])"
      ]
     },
     "metadata": {},
     "execution_count": 19
    }
   ]
  },
  {
   "cell_type": "code",
   "source": [
    "#What about using only numpy to pre-process?\n",
    "# https://stackoverflow.com/questions/66353849/confusions-about-the-input-shape-in-lstm-models\n",
    "\n",
    "n_features= 8 # how many predictors/Xs/features we have to predict y\n",
    "\n",
    "def gen_np(df, window):\n",
    "  num_windows = len(df) - window + 1\n",
    "  # make slices of data\n",
    "  Xs, ys = [], []\n",
    "\n",
    "  for w in range(num_windows):\n",
    "      current_slice = df.iloc[w:w+window]\n",
    "      X = current_slice[['snowtemp', 'soiltemp', 'omega', 'Tbh', 'Tbv', 'cf', 'h', 'voc']].values\n",
    "      y = current_slice['density'].values[-1]\n",
    "      Xs.append(X)\n",
    "      ys.append(y)\n",
    "\n",
    "  X = np.stack(Xs)\n",
    "  y = np.stack(ys)\n",
    "  return X, y\n",
    "\n",
    "X, y = gen_np(train1, n_input)\n",
    "\n",
    "# check the shape\n",
    "print(X.shape)\n",
    "print(y.shape)\n",
    "print(train1.head()) #for reference\n",
    "print(X[:5])\n",
    "print(y)"
   ],
   "metadata": {
    "id": "fWfv8PBHQrmk",
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 617
    },
    "outputId": "db416cf0-b605-467c-8ade-9aaccd8fed00",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "(3, 3, 8)\n",
      "(3,)\n"
     ]
    },
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "           density  snowtemp  soiltemp  omega       Tbh       Tbv        cf  \\\n",
       "13562138  0.412166  0.816258  0.998811    0.7  0.668379  0.677059  0.319354   \n",
       "13563663  0.417339  0.805035  0.996234    0.7  0.678995  0.684643  0.319354   \n",
       "13565561  0.428953  0.870483  0.999304    0.7  0.677228  0.677272  0.319354   \n",
       "13566803  0.436613  0.892357  0.993725    0.7  0.677911  0.682764  0.319354   \n",
       "13568994  0.461971  0.903925  0.998780    0.7  0.672902  0.674774  0.319354   \n",
       "\n",
       "                 h       voc       lat       lon       date  \n",
       "13562138  0.133275  0.703391 -54.73447 -68.10685 2019-08-11  \n",
       "13563663  0.133275  0.703552 -54.73447 -68.10685 2019-08-12  \n",
       "13565561  0.133275  0.703702 -54.73447 -68.10685 2019-08-14  \n",
       "13566803  0.133275  0.703682 -54.73447 -68.10685 2019-08-15  \n",
       "13568994  0.133275  0.703603 -54.73447 -68.10685 2019-08-19  "
      ],
      "text/html": [
       "\n",
       "  <div id=\"df-9a61ce5b-f085-485a-98d2-9cab5839fa14\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>density</th>\n",
       "      <th>snowtemp</th>\n",
       "      <th>soiltemp</th>\n",
       "      <th>omega</th>\n",
       "      <th>Tbh</th>\n",
       "      <th>Tbv</th>\n",
       "      <th>cf</th>\n",
       "      <th>h</th>\n",
       "      <th>voc</th>\n",
       "      <th>lat</th>\n",
       "      <th>lon</th>\n",
       "      <th>date</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>13562138</th>\n",
       "      <td>0.412166</td>\n",
       "      <td>0.816258</td>\n",
       "      <td>0.998811</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.668379</td>\n",
       "      <td>0.677059</td>\n",
       "      <td>0.319354</td>\n",
       "      <td>0.133275</td>\n",
       "      <td>0.703391</td>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-68.10685</td>\n",
       "      <td>2019-08-11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13563663</th>\n",
       "      <td>0.417339</td>\n",
       "      <td>0.805035</td>\n",
       "      <td>0.996234</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.678995</td>\n",
       "      <td>0.684643</td>\n",
       "      <td>0.319354</td>\n",
       "      <td>0.133275</td>\n",
       "      <td>0.703552</td>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-68.10685</td>\n",
       "      <td>2019-08-12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13565561</th>\n",
       "      <td>0.428953</td>\n",
       "      <td>0.870483</td>\n",
       "      <td>0.999304</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.677228</td>\n",
       "      <td>0.677272</td>\n",
       "      <td>0.319354</td>\n",
       "      <td>0.133275</td>\n",
       "      <td>0.703702</td>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-68.10685</td>\n",
       "      <td>2019-08-14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13566803</th>\n",
       "      <td>0.436613</td>\n",
       "      <td>0.892357</td>\n",
       "      <td>0.993725</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.677911</td>\n",
       "      <td>0.682764</td>\n",
       "      <td>0.319354</td>\n",
       "      <td>0.133275</td>\n",
       "      <td>0.703682</td>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-68.10685</td>\n",
       "      <td>2019-08-15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13568994</th>\n",
       "      <td>0.461971</td>\n",
       "      <td>0.903925</td>\n",
       "      <td>0.998780</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.672902</td>\n",
       "      <td>0.674774</td>\n",
       "      <td>0.319354</td>\n",
       "      <td>0.133275</td>\n",
       "      <td>0.703603</td>\n",
       "      <td>-54.73447</td>\n",
       "      <td>-68.10685</td>\n",
       "      <td>2019-08-19</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-9a61ce5b-f085-485a-98d2-9cab5839fa14')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-9a61ce5b-f085-485a-98d2-9cab5839fa14 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-9a61ce5b-f085-485a-98d2-9cab5839fa14');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ]
     },
     "metadata": {}
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "[[[0.81625822 0.99881078 0.69999972 0.66837909 0.67705922 0.3193543\n",
      "   0.13327522 0.70339097]\n",
      "  [0.80503465 0.99623414 0.69999972 0.67899542 0.68464269 0.3193543\n",
      "   0.13327522 0.70355208]\n",
      "  [0.87048334 0.99930381 0.69999972 0.67722763 0.67727244 0.3193543\n",
      "   0.13327522 0.7037019 ]]\n",
      "\n",
      " [[0.80503465 0.99623414 0.69999972 0.67899542 0.68464269 0.3193543\n",
      "   0.13327522 0.70355208]\n",
      "  [0.87048334 0.99930381 0.69999972 0.67722763 0.67727244 0.3193543\n",
      "   0.13327522 0.7037019 ]\n",
      "  [0.89235672 0.99372522 0.69999972 0.6779113  0.68276368 0.3193543\n",
      "   0.13327522 0.70368185]]\n",
      "\n",
      " [[0.87048334 0.99930381 0.69999972 0.67722763 0.67727244 0.3193543\n",
      "   0.13327522 0.7037019 ]\n",
      "  [0.89235672 0.99372522 0.69999972 0.6779113  0.68276368 0.3193543\n",
      "   0.13327522 0.70368185]\n",
      "  [0.90392522 0.99878022 0.69999972 0.67290206 0.67477389 0.3193543\n",
      "   0.13327522 0.70360281]]]\n",
      "[0.4289531  0.43661336 0.46197081]\n"
     ]
    }
   ]
  },
  {
   "cell_type": "code",
   "source": [
    "#Try appending np values: \n",
    "X2, y2 = gen_np(train2, n_input)\n",
    "print(\"train2:\")\n",
    "print(X2.shape)\n",
    "print(y2.shape)\n",
    "X = np.vstack([X, X2])\n",
    "y = np.concatenate([y, y2])\n",
    "print(\"joined 1 and 2:\")\n",
    "print(X.shape)\n",
    "print(y.shape)"
   ],
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "z0BJgefdX4J_",
    "outputId": "4a766a4c-216c-4731-99eb-b93b0213b4c8",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "train2:\n",
      "(3, 3, 8)\n",
      "(3,)\n",
      "joined 1 and 2:\n",
      "(6, 3, 8)\n",
      "(6,)\n"
     ]
    }
   ]
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Example model set up and test"
   ],
   "metadata": {
    "id": "mxWrLtAEQoq1",
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "source": [
    "#Initialize:\n",
    "def model_lstm(lrn_rate = 0.01, mntum = 0):#These variables do nothing??? Why even have them???\n",
    "  model = keras.models.Sequential()\n",
    "  #1st layer:\n",
    "  model.add(keras.layers.LSTM(50, activation='relu', return_sequences=True, input_shape=(n_input, n_features)))\n",
    "  model.add(keras.layers.Dropout(0.2))\n",
    "  #2nd layer:\n",
    "  model.add(keras.layers.LSTM(50))\n",
    "  model.add(keras.layers.Dropout(0.2))\n",
    "  #Output layer:\n",
    "  model.add(keras.layers.Dense(1))\n",
    "  model.compile(optimizer='adam', loss='mean_squared_error')\n",
    "  #model.summary()\n",
    "  return(model)"
   ],
   "metadata": {
    "id": "f8A21VPCOqXC",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": []
  },
  {
   "cell_type": "code",
   "source": [
    "model = model_lstm(lrn_rate = 0.01, mntum = 0)"
   ],
   "metadata": {
    "id": "I81Jzhl1TK3l",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": []
  },
  {
   "cell_type": "code",
   "source": [
    "model.summary()"
   ],
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "0yc5oQuWUFdZ",
    "outputId": "5d3cf542-a0ee-4c80-d43d-95799b1141f8",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " lstm_4 (LSTM)               (None, 3, 50)             12000     \n",
      "                                                                 \n",
      " dropout_4 (Dropout)         (None, 3, 50)             0         \n",
      "                                                                 \n",
      " lstm_5 (LSTM)               (None, 50)                20200     \n",
      "                                                                 \n",
      " dropout_5 (Dropout)         (None, 50)                0         \n",
      "                                                                 \n",
      " dense_2 (Dense)             (None, 1)                 51        \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 32,251\n",
      "Trainable params: 32,251\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ]
  },
  {
   "cell_type": "code",
   "source": [
    "#modelfit = model.fit(x = generator[0][0], y=generator.targets[1:-(n_input-1)], epochs = 5) #works!\n",
    "#modelfit = model.fit(x = gen, y=ygen, epochs = 5) #Works!\n",
    "modelfit = model.fit(x = X, y = y, epochs = 5) #works"
   ],
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "PCRWt_d_T7Kn",
    "outputId": "a96233f1-3968-4062-89c0-06a06971bd0f",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Epoch 1/5\n",
      "1/1 [==============================] - 4s 4s/step - loss: 0.2085\n",
      "Epoch 2/5\n",
      "1/1 [==============================] - 0s 16ms/step - loss: 0.1824\n",
      "Epoch 3/5\n",
      "1/1 [==============================] - 0s 12ms/step - loss: 0.1728\n",
      "Epoch 4/5\n",
      "1/1 [==============================] - 0s 12ms/step - loss: 0.1551\n",
      "Epoch 5/5\n",
      "1/1 [==============================] - 0s 12ms/step - loss: 0.1408\n"
     ]
    }
   ]
  },
  {
   "cell_type": "code",
   "source": [
    "loss_per_epoch = model.history.history['loss']\n",
    "plt.plot(range(len(loss_per_epoch)),loss_per_epoch)"
   ],
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 283
    },
    "id": "bHHrNWpHUfPQ",
    "outputId": "de005351-688f-4dec-d642-9db6305358f0",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7fdb519a3340>]"
      ]
     },
     "metadata": {},
     "execution_count": 45
    },
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ],
      "image/png": "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\n"
     },
     "metadata": {
      "needs_background": "light"
     }
    }
   ]
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Example test set prep\n"
   ],
   "metadata": {
    "id": "icZfPYef8NdK",
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "source": [
    "test = pd.read_parquet('test_scaled_LatLonDateUnscaled.parquet')\n",
    "\n",
    "print(test.columns)\n",
    "\n",
    "print(test['date'].max())\n",
    "print(test['date'].min())\n",
    "print(len(test))\n",
    "print(test.info())\n",
    "test.head()"
   ],
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 671
    },
    "id": "lORabcV78TBX",
    "outputId": "32430c19-06c7-47f1-e012-e283e6ae3205",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": 7,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Index(['density', 'snowtemp', 'soiltemp', 'omega', 'Tbh', 'Tbv', 'cf', 'h',\n",
      "       'voc', 'lat', 'lon', 'date'],\n",
      "      dtype='object')\n",
      "2019-12-31 00:00:00\n",
      "2019-01-01 00:00:00\n",
      "18329587\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 18329587 entries, 0 to 18329586\n",
      "Data columns (total 12 columns):\n",
      " #   Column    Dtype         \n",
      "---  ------    -----         \n",
      " 0   density   float64       \n",
      " 1   snowtemp  float64       \n",
      " 2   soiltemp  float64       \n",
      " 3   omega     float64       \n",
      " 4   Tbh       float64       \n",
      " 5   Tbv       float64       \n",
      " 6   cf        float64       \n",
      " 7   h         float64       \n",
      " 8   voc       float64       \n",
      " 9   lat       float64       \n",
      " 10  lon       float64       \n",
      " 11  date      datetime64[ns]\n",
      "dtypes: datetime64[ns](1), float64(11)\n",
      "memory usage: 1.8 GB\n",
      "None\n"
     ]
    },
    {
     "data": {
      "text/plain": "    density  snowtemp  soiltemp     omega       Tbh       Tbv        cf  \\\n0  0.166630  0.908916  0.998302  0.779338  0.710750  0.708752  0.305672   \n1  0.176869  0.955726  0.995804  0.783470  0.715650  0.708749  0.311526   \n2  0.180525  0.959748  0.996218  0.790908  0.722035  0.714708  0.322348   \n3  0.183057  0.966050  0.996963  0.790081  0.721543  0.716810  0.302173   \n4  0.167732  0.926476  0.995828  0.789999  0.709735  0.705815  0.303099   \n\n          h       voc        lat        lon       date  \n0  0.124832  0.240164  62.684110 -160.44087 2019-01-01  \n1  0.123756  0.228918  62.532246 -160.44087 2019-01-01  \n2  0.123569  0.242246  62.381150 -160.44087 2019-01-01  \n3  0.123513  0.240669  62.230810 -160.44087 2019-01-01  \n4  0.123518  0.239886  62.684110 -160.34750 2019-01-01  ",
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>density</th>\n      <th>snowtemp</th>\n      <th>soiltemp</th>\n      <th>omega</th>\n      <th>Tbh</th>\n      <th>Tbv</th>\n      <th>cf</th>\n      <th>h</th>\n      <th>voc</th>\n      <th>lat</th>\n      <th>lon</th>\n      <th>date</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>0.166630</td>\n      <td>0.908916</td>\n      <td>0.998302</td>\n      <td>0.779338</td>\n      <td>0.710750</td>\n      <td>0.708752</td>\n      <td>0.305672</td>\n      <td>0.124832</td>\n      <td>0.240164</td>\n      <td>62.684110</td>\n      <td>-160.44087</td>\n      <td>2019-01-01</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>0.176869</td>\n      <td>0.955726</td>\n      <td>0.995804</td>\n      <td>0.783470</td>\n      <td>0.715650</td>\n      <td>0.708749</td>\n      <td>0.311526</td>\n      <td>0.123756</td>\n      <td>0.228918</td>\n      <td>62.532246</td>\n      <td>-160.44087</td>\n      <td>2019-01-01</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>0.180525</td>\n      <td>0.959748</td>\n      <td>0.996218</td>\n      <td>0.790908</td>\n      <td>0.722035</td>\n      <td>0.714708</td>\n      <td>0.322348</td>\n      <td>0.123569</td>\n      <td>0.242246</td>\n      <td>62.381150</td>\n      <td>-160.44087</td>\n      <td>2019-01-01</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>0.183057</td>\n      <td>0.966050</td>\n      <td>0.996963</td>\n      <td>0.790081</td>\n      <td>0.721543</td>\n      <td>0.716810</td>\n      <td>0.302173</td>\n      <td>0.123513</td>\n      <td>0.240669</td>\n      <td>62.230810</td>\n      <td>-160.44087</td>\n      <td>2019-01-01</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>0.167732</td>\n      <td>0.926476</td>\n      <td>0.995828</td>\n      <td>0.789999</td>\n      <td>0.709735</td>\n      <td>0.705815</td>\n      <td>0.303099</td>\n      <td>0.123518</td>\n      <td>0.239886</td>\n      <td>62.684110</td>\n      <td>-160.34750</td>\n      <td>2019-01-01</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ]
  },
  {
   "cell_type": "code",
   "source": [
    "test1 = test[(test['lat'] == latlon['lat'].iloc[0]) & (test['lon'] == latlon['lon'].iloc[0])]\n",
    "print(len(test1))\n",
    "print(test1['date'].min())\n",
    "print(test1['date'].max())\n",
    "test1.head()\n",
    "#Testing with the same lat/lon as train, this won't be matched in the larger set."
   ],
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 260
    },
    "id": "Z4UZ0JkB_Szh",
    "outputId": "9299d5b0-2344-48e9-c1ca-bbab43356371",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "72\n",
      "2020-01-01 00:00:00\n",
      "2020-11-20 00:00:00\n"
     ]
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "           density  snowtemp  soiltemp     omega       Tbh       Tbv  \\\n",
       "18329763  0.233223  0.537159  0.959042  0.779338  0.707065  0.712970   \n",
       "18517417  0.215192  0.482522  0.962540  0.779338  0.709041  0.716228   \n",
       "18577061  0.217689  0.420847  0.960800  0.779338  0.705201  0.705174   \n",
       "18674566  0.219739  0.384268  0.954615  0.779338  0.700979  0.706267   \n",
       "18772980  0.221373  0.323696  0.944928  0.779338  0.705278  0.709230   \n",
       "\n",
       "                cf         h       voc       lat        lon       date  \n",
       "18329763  0.305672  0.124832  0.234089  62.68411 -160.44087 2020-01-01  \n",
       "18517417  0.305672  0.124832  0.234037  62.68411 -160.44087 2020-01-03  \n",
       "18577061  0.305672  0.124832  0.234016  62.68411 -160.44087 2020-01-04  \n",
       "18674566  0.305672  0.124832  0.233997  62.68411 -160.44087 2020-01-05  \n",
       "18772980  0.305672  0.124832  0.233980  62.68411 -160.44087 2020-01-06  "
      ],
      "text/html": [
       "\n",
       "  <div id=\"df-51af95ce-04a3-4554-baf4-bbf797d07466\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>density</th>\n",
       "      <th>snowtemp</th>\n",
       "      <th>soiltemp</th>\n",
       "      <th>omega</th>\n",
       "      <th>Tbh</th>\n",
       "      <th>Tbv</th>\n",
       "      <th>cf</th>\n",
       "      <th>h</th>\n",
       "      <th>voc</th>\n",
       "      <th>lat</th>\n",
       "      <th>lon</th>\n",
       "      <th>date</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>18329763</th>\n",
       "      <td>0.233223</td>\n",
       "      <td>0.537159</td>\n",
       "      <td>0.959042</td>\n",
       "      <td>0.779338</td>\n",
       "      <td>0.707065</td>\n",
       "      <td>0.712970</td>\n",
       "      <td>0.305672</td>\n",
       "      <td>0.124832</td>\n",
       "      <td>0.234089</td>\n",
       "      <td>62.68411</td>\n",
       "      <td>-160.44087</td>\n",
       "      <td>2020-01-01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18517417</th>\n",
       "      <td>0.215192</td>\n",
       "      <td>0.482522</td>\n",
       "      <td>0.962540</td>\n",
       "      <td>0.779338</td>\n",
       "      <td>0.709041</td>\n",
       "      <td>0.716228</td>\n",
       "      <td>0.305672</td>\n",
       "      <td>0.124832</td>\n",
       "      <td>0.234037</td>\n",
       "      <td>62.68411</td>\n",
       "      <td>-160.44087</td>\n",
       "      <td>2020-01-03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18577061</th>\n",
       "      <td>0.217689</td>\n",
       "      <td>0.420847</td>\n",
       "      <td>0.960800</td>\n",
       "      <td>0.779338</td>\n",
       "      <td>0.705201</td>\n",
       "      <td>0.705174</td>\n",
       "      <td>0.305672</td>\n",
       "      <td>0.124832</td>\n",
       "      <td>0.234016</td>\n",
       "      <td>62.68411</td>\n",
       "      <td>-160.44087</td>\n",
       "      <td>2020-01-04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18674566</th>\n",
       "      <td>0.219739</td>\n",
       "      <td>0.384268</td>\n",
       "      <td>0.954615</td>\n",
       "      <td>0.779338</td>\n",
       "      <td>0.700979</td>\n",
       "      <td>0.706267</td>\n",
       "      <td>0.305672</td>\n",
       "      <td>0.124832</td>\n",
       "      <td>0.233997</td>\n",
       "      <td>62.68411</td>\n",
       "      <td>-160.44087</td>\n",
       "      <td>2020-01-05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18772980</th>\n",
       "      <td>0.221373</td>\n",
       "      <td>0.323696</td>\n",
       "      <td>0.944928</td>\n",
       "      <td>0.779338</td>\n",
       "      <td>0.705278</td>\n",
       "      <td>0.709230</td>\n",
       "      <td>0.305672</td>\n",
       "      <td>0.124832</td>\n",
       "      <td>0.233980</td>\n",
       "      <td>62.68411</td>\n",
       "      <td>-160.44087</td>\n",
       "      <td>2020-01-06</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-51af95ce-04a3-4554-baf4-bbf797d07466')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-51af95ce-04a3-4554-baf4-bbf797d07466 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-51af95ce-04a3-4554-baf4-bbf797d07466');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ]
     },
     "metadata": {},
     "execution_count": 95
    }
   ]
  },
  {
   "cell_type": "code",
   "source": [
    "x_test1 = test1[['snowtemp', 'soiltemp', 'omega', 'Tbh', 'Tbv', 'cf', 'h', 'voc']]\n",
    "x_test1 = np.array(x_test1)\n",
    "y_test1 = test1[['density']]\n",
    "y_test1 = np.array(y_test1)\n",
    "y_test1 = y_test1.reshape(-1) #now is values in one array, not an array of arrays size 1. \n",
    "\n",
    "print(x_test1.shape)\n",
    "print(y_test1.shape)"
   ],
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "QeuCylhGCjMz",
    "outputId": "157ea7a6-508e-4ed1-eed7-0c181ff58610",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "(72, 8)\n",
      "(72,)\n"
     ]
    }
   ]
  },
  {
   "cell_type": "code",
   "source": [
    "X_test, y_test = gen_np(test1, n_input)"
   ],
   "metadata": {
    "id": "f_rV9TFkjZYp",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": []
  },
  {
   "cell_type": "code",
   "source": [
    "test_generator = TimeseriesGenerator(x_test1, np.zeros(len(x_test1)), length=n_input, batch_size=b_size)\n",
    "print(test_generator[0][0].shape) #We get the full 50 samples, could have had more since the actual lenth of test is 71."
   ],
   "metadata": {
    "id": "FXGDSeVpU8p0",
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "outputId": "c517fa7e-cc87-4819-e7c0-934e643d9083",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "(50, 3, 8)\n"
     ]
    }
   ]
  },
  {
   "cell_type": "code",
   "source": [
    "#y_hat = model.predict(test_generator)\n",
    "y_hat = model.predict(X_test)"
   ],
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "er447vhBjzfc",
    "outputId": "3c4c5d3c-b74f-47ef-8125-e40e483ef096",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "3/3 [==============================] - 0s 5ms/step\n"
     ]
    }
   ]
  },
  {
   "cell_type": "code",
   "source": [
    "#y_pred = Yscaler.inverse_transform(y_pred_scaled)\n",
    "#results = pd.DataFrame({'y_test1':test1['density'].values[n_input:],'y_hat':y_hat.ravel()})\n",
    "results = pd.DataFrame({'y_test1':y_test,'y_hat':y_hat.ravel()})\n",
    "print(results)"
   ],
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "Z9mF-NljEME4",
    "outputId": "156e633b-eec4-4736-aa2a-1c11244cd152",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "     y_test1     y_hat\n",
      "0   0.217689  0.235305\n",
      "1   0.219739  0.233692\n",
      "2   0.221373  0.232147\n",
      "3   0.222658  0.231271\n",
      "4   0.223785  0.230292\n",
      "..       ...       ...\n",
      "65  0.142432  0.245259\n",
      "66  0.182735  0.240560\n",
      "67  0.183829  0.239973\n",
      "68  0.182399  0.239260\n",
      "69  0.177009  0.241614\n",
      "\n",
      "[70 rows x 2 columns]\n"
     ]
    }
   ]
  },
  {
   "cell_type": "code",
   "source": [
    "print('R2', skm.r2_score(results['y_test1'], results['y_hat']))\n",
    "print('MAPE', skm.mean_absolute_percentage_error(results['y_test1'], results['y_hat']))\n",
    "print('MAE', skm.mean_absolute_error(results['y_test1'], results['y_hat']))\n",
    "print('RMSE', skm.mean_squared_error(results['y_test1'], results['y_hat'], squared = False))\n",
    "print( 'MSE', skm.mean_squared_error(results['y_test1'], results['y_hat']))"
   ],
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "DxhI8hAnIRI2",
    "outputId": "c67c590a-f6de-4558-95ef-4d79787288b7",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": null,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "R2 -0.018116347457838078\n",
      "MAPE 0.13819738284814045\n",
      "MAE 0.033359309261546344\n",
      "RMSE 0.043977989876406685\n",
      "MSE 0.001934063593569329\n"
     ]
    }
   ]
  },
  {
   "cell_type": "markdown",
   "source": [
    "Old:\n",
    "R2 -17.570187310639216\n",
    "MAPE 0.7333341272085523\n",
    "MAE 0.1834174423880057\n",
    "RMSE 0.1885665896580615\n",
    "MSE 0.035557358735271744"
   ],
   "metadata": {
    "id": "eocECDFnkBrU",
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Data Prep"
   ],
   "metadata": {
    "id": "prHcypSQHaXr",
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "source": [
    "#Peramaeters: \n",
    "n_input = 60 #how many samples/rows/timesteps to look in the past in order to forecast the next sample\n",
    "n_features = 7 # how many predictors/Xs/features we have to predict y"
   ],
   "metadata": {
    "id": "UeX2kmu2IT5s",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": 25,
   "outputs": []
  },
  {
   "cell_type": "code",
   "source": [
    "# Import data:\n",
    "train = pd.read_parquet('train_scaled_LatLonDateUnscaled.parquet')\n",
    "\n",
    "print(train.columns)\n",
    "\n",
    "print(train['date'].max())\n",
    "print(train['date'].min())\n",
    "print(len(train))\n",
    "print(train.info())\n",
    "print(train.shape)\n",
    "train.head()"
   ],
   "metadata": {
    "id": "az25ZGQhHen0",
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 689
    },
    "outputId": "b69ecfe6-0514-45a0-db17-2ff3bf2df771",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": 35,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Index(['density', 'snowtemp', 'soiltemp', 'omega', 'Tbh', 'Tbv', 'cf', 'h',\n",
      "       'voc', 'lat', 'lon', 'date'],\n",
      "      dtype='object')\n",
      "2019-12-31 00:00:00\n",
      "2019-01-01 00:00:00\n",
      "18329587\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 18329587 entries, 0 to 18329586\n",
      "Data columns (total 12 columns):\n",
      " #   Column    Dtype         \n",
      "---  ------    -----         \n",
      " 0   density   float64       \n",
      " 1   snowtemp  float64       \n",
      " 2   soiltemp  float64       \n",
      " 3   omega     float64       \n",
      " 4   Tbh       float64       \n",
      " 5   Tbv       float64       \n",
      " 6   cf        float64       \n",
      " 7   h         float64       \n",
      " 8   voc       float64       \n",
      " 9   lat       float64       \n",
      " 10  lon       float64       \n",
      " 11  date      datetime64[ns]\n",
      "dtypes: datetime64[ns](1), float64(11)\n",
      "memory usage: 1.8 GB\n",
      "None\n",
      "(18329587, 12)\n"
     ]
    },
    {
     "data": {
      "text/plain": "    density  snowtemp  soiltemp     omega       Tbh       Tbv        cf  \\\n0  0.166630  0.908916  0.998302  0.779338  0.710750  0.708752  0.305672   \n1  0.176869  0.955726  0.995804  0.783470  0.715650  0.708749  0.311526   \n2  0.180525  0.959748  0.996218  0.790908  0.722035  0.714708  0.322348   \n3  0.183057  0.966050  0.996963  0.790081  0.721543  0.716810  0.302173   \n4  0.167732  0.926476  0.995828  0.789999  0.709735  0.705815  0.303099   \n\n          h       voc        lat        lon       date  \n0  0.124832  0.240164  62.684110 -160.44087 2019-01-01  \n1  0.123756  0.228918  62.532246 -160.44087 2019-01-01  \n2  0.123569  0.242246  62.381150 -160.44087 2019-01-01  \n3  0.123513  0.240669  62.230810 -160.44087 2019-01-01  \n4  0.123518  0.239886  62.684110 -160.34750 2019-01-01  ",
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>density</th>\n      <th>snowtemp</th>\n      <th>soiltemp</th>\n      <th>omega</th>\n      <th>Tbh</th>\n      <th>Tbv</th>\n      <th>cf</th>\n      <th>h</th>\n      <th>voc</th>\n      <th>lat</th>\n      <th>lon</th>\n      <th>date</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>0.166630</td>\n      <td>0.908916</td>\n      <td>0.998302</td>\n      <td>0.779338</td>\n      <td>0.710750</td>\n      <td>0.708752</td>\n      <td>0.305672</td>\n      <td>0.124832</td>\n      <td>0.240164</td>\n      <td>62.684110</td>\n      <td>-160.44087</td>\n      <td>2019-01-01</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>0.176869</td>\n      <td>0.955726</td>\n      <td>0.995804</td>\n      <td>0.783470</td>\n      <td>0.715650</td>\n      <td>0.708749</td>\n      <td>0.311526</td>\n      <td>0.123756</td>\n      <td>0.228918</td>\n      <td>62.532246</td>\n      <td>-160.44087</td>\n      <td>2019-01-01</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>0.180525</td>\n      <td>0.959748</td>\n      <td>0.996218</td>\n      <td>0.790908</td>\n      <td>0.722035</td>\n      <td>0.714708</td>\n      <td>0.322348</td>\n      <td>0.123569</td>\n      <td>0.242246</td>\n      <td>62.381150</td>\n      <td>-160.44087</td>\n      <td>2019-01-01</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>0.183057</td>\n      <td>0.966050</td>\n      <td>0.996963</td>\n      <td>0.790081</td>\n      <td>0.721543</td>\n      <td>0.716810</td>\n      <td>0.302173</td>\n      <td>0.123513</td>\n      <td>0.240669</td>\n      <td>62.230810</td>\n      <td>-160.44087</td>\n      <td>2019-01-01</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>0.167732</td>\n      <td>0.926476</td>\n      <td>0.995828</td>\n      <td>0.789999</td>\n      <td>0.709735</td>\n      <td>0.705815</td>\n      <td>0.303099</td>\n      <td>0.123518</td>\n      <td>0.239886</td>\n      <td>62.684110</td>\n      <td>-160.34750</td>\n      <td>2019-01-01</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "123873\n"
     ]
    }
   ],
   "source": [
    "latlon = latlon_setup(train, n_input)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "outputs": [
    {
     "data": {
      "text/plain": "<matplotlib.patches.StepPatch at 0x1cf738e4040>"
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": "<Figure size 640x480 with 1 Axes>",
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAkIAAAGdCAYAAAD+JxxnAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8o6BhiAAAACXBIWXMAAA9hAAAPYQGoP6dpAAA5uklEQVR4nO3df1RVdb7/8ReQ54A/DogGBwwRrTQVqWxkWFNaV65HL9PE1ExmTpGRpYOV4hSX7/Jn0wqvrH6P5TQzZbNqSl2rbEaduoi/phGxUPJHyVIHM4sDMybnKCogfL5/zHJfT5hCQkfYz8daew17f97nc94fNnZec87eEGKMMQIAALCh0GA3AAAAECwEIQAAYFsEIQAAYFsEIQAAYFsEIQAAYFsEIQAAYFsEIQAAYFsEIQAAYFuXBbuBS1lzc7O++uor9erVSyEhIcFuBwAAtIIxRseOHVN8fLxCQ8//ng9B6Dy++uorJSQkBLsNAADwHXzxxRe64oorzltDEDqPXr16Sfr3N9LlcgW5GwAA0Bp+v18JCQnW6/j5EITO48zHYS6XiyAEAEAn05rLWrhYGgAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZ/fT6Ivqw9qaN1DcFuo01693CoX1REsNsAAKBdEISC5Mvak0p/epNONjYFu5U2iegWpnWzxxCGAABdAkEoSI7WNehkY5Oem3itrozpGex2WmV/zXHNXF6uo3UNBCEAQJdAEAqyK2N6ani/yGC3AQCALXGxNAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsK02BaGCggL94Ac/UK9evRQTE6PMzExVVFQE1Jw6dUo5OTnq06ePevbsqTvuuEPV1dUBNYcOHVJGRoa6d++umJgYPfbYYzp9+nRAzcaNG3X99dfL6XTqyiuv1LJly1r0s2TJEg0YMEDh4eFKTU3Vtm3b2twLAACwrzYFoU2bNiknJ0dbt25VUVGRGhsbNW7cONXV1Vk1s2bN0l/+8hetXLlSmzZt0ldffaXbb7/dGm9qalJGRoYaGhq0ZcsWvf7661q2bJnmzZtn1VRWViojI0O33HKLysvLNXPmTD3wwAP64IMPrJrly5crNzdX8+fP1/bt25WSkiKPx6OamppW9wIAAGzOXISamhojyWzatMkYY0xtba3p1q2bWblypVXz2WefGUmmpKTEGGPM2rVrTWhoqPF6vVbNyy+/bFwul6mvrzfGGPP444+bYcOGBTzXxIkTjcfjsfZHjRplcnJyrP2mpiYTHx9vCgoKWt3Lhfh8PiPJ+Hy+VtW3xa7DtSYxb7XZdbi23efuKJ2xZwCA/bTl9fuirhHy+XySpOjoaElSWVmZGhsblZ6ebtUMGTJE/fv3V0lJiSSppKREycnJio2NtWo8Ho/8fr/27Nlj1Zw9x5maM3M0NDSorKwsoCY0NFTp6elWTWt6+ab6+nr5/f6ADQAAdF3fOQg1Nzdr5syZ+tGPfqThw4dLkrxerxwOh6KiogJqY2Nj5fV6rZqzQ9CZ8TNj56vx+/06efKk/vWvf6mpqemcNWfPcaFevqmgoECRkZHWlpCQ0MrvBgAA6Iy+cxDKycnR7t279fbbb7dnP0GVn58vn89nbV988UWwWwIAAB3oO/31+RkzZmj16tXavHmzrrjiCuu42+1WQ0ODamtrA96Jqa6ultvttmq+eXfXmTu5zq755t1d1dXVcrlcioiIUFhYmMLCws5Zc/YcF+rlm5xOp5xOZxu+EwAAoDNr0ztCxhjNmDFD7777rtavX6+kpKSA8ZEjR6pbt24qLi62jlVUVOjQoUNKS0uTJKWlpWnXrl0Bd3cVFRXJ5XJp6NChVs3Zc5ypOTOHw+HQyJEjA2qam5tVXFxs1bSmFwAAYG9tekcoJydHf/rTn/Tee++pV69e1rU2kZGRioiIUGRkpLKzs5Wbm6vo6Gi5XC49/PDDSktL0w9/+ENJ0rhx4zR06FDdc889Wrx4sbxer+bMmaOcnBzr3Zhp06bpN7/5jR5//HHdf//9Wr9+vVasWKE1a9ZYveTm5iorK0s33HCDRo0apeeee051dXWaMmWK1dOFegEAADbXltvRJJ1ze+2116yakydPml/+8pemd+/epnv37uanP/2pqaqqCpjn4MGDZsKECSYiIsL07dvXzJ492zQ2NgbUbNiwwVx77bXG4XCYgQMHBjzHGS+++KLp37+/cTgcZtSoUWbr1q0B463p5Xy4fT5QZ+wZAGA/bXn9DjHGmODFsEub3+9XZGSkfD6fXC5Xu869+0uffvzih1r98I0a3i+yXefuKJ2xZwCA/bTl9Zu/NQYAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyLIAQAAGyrzUFo8+bNuvXWWxUfH6+QkBCtWrUqYDwkJOScW2FhoVUzYMCAFuOLFi0KmGfnzp266aabFB4eroSEBC1evLhFLytXrtSQIUMUHh6u5ORkrV27NmDcGKN58+YpLi5OERERSk9P1759+9q6ZAAA0EW1OQjV1dUpJSVFS5YsOed4VVVVwPbqq68qJCREd9xxR0DdE088EVD38MMPW2N+v1/jxo1TYmKiysrKVFhYqAULFuiVV16xarZs2aJJkyYpOztbO3bsUGZmpjIzM7V7926rZvHixXrhhRe0dOlSlZaWqkePHvJ4PDp16lRblw0AALqgy9r6gAkTJmjChAnfOu52uwP233vvPd1yyy0aOHBgwPFevXq1qD3jzTffVENDg1599VU5HA4NGzZM5eXleuaZZ/Tggw9Kkp5//nmNHz9ejz32mCTp17/+tYqKivSb3/xGS5culTFGzz33nObMmaPbbrtNkvTHP/5RsbGxWrVqle666662Lh0AAHQxHXqNUHV1tdasWaPs7OwWY4sWLVKfPn103XXXqbCwUKdPn7bGSkpKNHr0aDkcDuuYx+NRRUWFjh49atWkp6cHzOnxeFRSUiJJqqyslNfrDaiJjIxUamqqVfNN9fX18vv9ARsAAOi62vyOUFu8/vrr6tWrl26//faA44888oiuv/56RUdHa8uWLcrPz1dVVZWeeeYZSZLX61VSUlLAY2JjY62x3r17y+v1WsfOrvF6vVbd2Y87V803FRQUaOHChd9xtQAAoLPp0CD06quvavLkyQoPDw84npuba309YsQIORwOPfTQQyooKJDT6ezIls4rPz8/oDe/36+EhISg9QMAADpWh3009re//U0VFRV64IEHLlibmpqq06dP6+DBg5L+fZ1RdXV1QM2Z/TPXFX1bzdnjZz/uXDXf5HQ65XK5AjYAANB1dVgQ+sMf/qCRI0cqJSXlgrXl5eUKDQ1VTEyMJCktLU2bN29WY2OjVVNUVKTBgwerd+/eVk1xcXHAPEVFRUpLS5MkJSUlye12B9T4/X6VlpZaNQAAwN7a/NHY8ePHtX//fmu/srJS5eXlio6OVv/+/SX9O3CsXLlSTz/9dIvHl5SUqLS0VLfccot69eqlkpISzZo1S7/4xS+skHP33Xdr4cKFys7OVl5ennbv3q3nn39ezz77rDXPo48+qjFjxujpp59WRkaG3n77bX388cfWLfYhISGaOXOmnnzySV111VVKSkrS3LlzFR8fr8zMzLYuGwAAdEWmjTZs2GAktdiysrKsmt/+9rcmIiLC1NbWtnh8WVmZSU1NNZGRkSY8PNxcc8015qmnnjKnTp0KqPvkk0/MjTfeaJxOp+nXr59ZtGhRi7lWrFhhrr76auNwOMywYcPMmjVrAsabm5vN3LlzTWxsrHE6nWbs2LGmoqKi1Wv1+XxGkvH5fK1+TGvtOlxrEvNWm12HW36PLlWdsWcAgP205fU7xBhjgpjDLml+v1+RkZHy+Xztfr3Q7i99+vGLH2r1wzdqeL/Idp27o3TGngEA9tOW12/+1hgAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALAtghAAALCtNgehzZs369Zbb1V8fLxCQkK0atWqgPH77rtPISEhAdv48eMDar7++mtNnjxZLpdLUVFRys7O1vHjxwNqdu7cqZtuuknh4eFKSEjQ4sWLW/SycuVKDRkyROHh4UpOTtbatWsDxo0xmjdvnuLi4hQREaH09HTt27evrUsGAABdVJuDUF1dnVJSUrRkyZJvrRk/fryqqqqs7a233goYnzx5svbs2aOioiKtXr1amzdv1oMPPmiN+/1+jRs3TomJiSorK1NhYaEWLFigV155xarZsmWLJk2apOzsbO3YsUOZmZnKzMzU7t27rZrFixfrhRde0NKlS1VaWqoePXrI4/Ho1KlTbV02AADoisxFkGTefffdgGNZWVnmtttu+9bHfPrpp0aS+eijj6xjf/3rX01ISIj58ssvjTHGvPTSS6Z3796mvr7eqsnLyzODBw+29u+8806TkZERMHdqaqp56KGHjDHGNDc3G7fbbQoLC63x2tpa43Q6zVtvvdWq9fl8PiPJ+Hy+VtW3xa7DtSYxb7XZdbi23efuKJ2xZwCA/bTl9btDrhHauHGjYmJiNHjwYE2fPl1HjhyxxkpKShQVFaUbbrjBOpaenq7Q0FCVlpZaNaNHj5bD4bBqPB6PKioqdPToUasmPT094Hk9Ho9KSkokSZWVlfJ6vQE1kZGRSk1NtWq+qb6+Xn6/P2ADAABdV7sHofHjx+uPf/yjiouL9T//8z/atGmTJkyYoKamJkmS1+tVTExMwGMuu+wyRUdHy+v1WjWxsbEBNWf2L1Rz9vjZjztXzTcVFBQoMjLS2hISEtq8fgAA0Hlc1t4T3nXXXdbXycnJGjFihAYNGqSNGzdq7Nix7f107So/P1+5ubnWvt/vJwwBANCFdfjt8wMHDlTfvn21f/9+SZLb7VZNTU1AzenTp/X111/L7XZbNdXV1QE1Z/YvVHP2+NmPO1fNNzmdTrlcroANAAB0XR0ehA4fPqwjR44oLi5OkpSWlqba2lqVlZVZNevXr1dzc7NSU1Otms2bN6uxsdGqKSoq0uDBg9W7d2+rpri4OOC5ioqKlJaWJklKSkqS2+0OqPH7/SotLbVqAACAvbU5CB0/flzl5eUqLy+X9O+LksvLy3Xo0CEdP35cjz32mLZu3aqDBw+quLhYt912m6688kp5PB5J0jXXXKPx48dr6tSp2rZtm/7+979rxowZuuuuuxQfHy9Juvvuu+VwOJSdna09e/Zo+fLlev755wM+tnr00Uf1/vvv6+mnn9bevXu1YMECffzxx5oxY4YkKSQkRDNnztSTTz6pP//5z9q1a5fuvfdexcfHKzMz8yK/bQAAoEto6y1pGzZsMJJabFlZWebEiRNm3Lhx5vLLLzfdunUziYmJZurUqcbr9QbMceTIETNp0iTTs2dP43K5zJQpU8yxY8cCaj755BNz4403GqfTafr162cWLVrUopcVK1aYq6++2jgcDjNs2DCzZs2agPHm5mYzd+5cExsba5xOpxk7dqypqKho9Vq5fT5QZ+wZAGA/bXn9DjHGmCDmsEua3+9XZGSkfD5fu18vtPtLn3784oda/fCNGt4vsl3n7iidsWcAgP205fWbvzUGAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsq81BaPPmzbr11lsVHx+vkJAQrVq1yhprbGxUXl6ekpOT1aNHD8XHx+vee+/VV199FTDHgAEDFBISErAtWrQooGbnzp266aabFB4eroSEBC1evLhFLytXrtSQIUMUHh6u5ORkrV27NmDcGKN58+YpLi5OERERSk9P1759+9q6ZAAA0EW1OQjV1dUpJSVFS5YsaTF24sQJbd++XXPnztX27dv1zjvvqKKiQj/5yU9a1D7xxBOqqqqytocfftga8/v9GjdunBITE1VWVqbCwkItWLBAr7zyilWzZcsWTZo0SdnZ2dqxY4cyMzOVmZmp3bt3WzWLFy/WCy+8oKVLl6q0tFQ9evSQx+PRqVOn2rpsAADQFZmLIMm8++67563Ztm2bkWQ+//xz61hiYqJ59tlnv/UxL730kundu7epr6+3juXl5ZnBgwdb+3feeafJyMgIeFxqaqp56KGHjDHGNDc3G7fbbQoLC63x2tpa43Q6zVtvvdWa5Rmfz2ckGZ/P16r6tth1uNYk5q02uw7XtvvcHaUz9gwAsJ+2vH53+DVCPp9PISEhioqKCji+aNEi9enTR9ddd50KCwt1+vRpa6ykpESjR4+Ww+Gwjnk8HlVUVOjo0aNWTXp6esCcHo9HJSUlkqTKykp5vd6AmsjISKWmplo1AADA3i7ryMlPnTqlvLw8TZo0SS6Xyzr+yCOP6Prrr1d0dLS2bNmi/Px8VVVV6ZlnnpEkeb1eJSUlBcwVGxtrjfXu3Vter9c6dnaN1+u16s5+3Llqvqm+vl719fXWvt/v/y7LBgAAnUSHBaHGxkbdeeedMsbo5ZdfDhjLzc21vh4xYoQcDoceeughFRQUyOl0dlRLF1RQUKCFCxcG7fkBAMD3q0M+GjsTgj7//HMVFRUFvBt0LqmpqTp9+rQOHjwoSXK73aqurg6oObPvdrvPW3P2+NmPO1fNN+Xn58vn81nbF1980YrVAgCAzqrdg9CZELRv3z6tW7dOffr0ueBjysvLFRoaqpiYGElSWlqaNm/erMbGRqumqKhIgwcPVu/eva2a4uLigHmKioqUlpYmSUpKSpLb7Q6o8fv9Ki0ttWq+yel0yuVyBWwAAKDravNHY8ePH9f+/fut/crKSpWXlys6OlpxcXH62c9+pu3bt2v16tVqamqyrseJjo6Ww+FQSUmJSktLdcstt6hXr14qKSnRrFmz9Itf/MIKOXfffbcWLlyo7Oxs5eXlaffu3Xr++ef17LPPWs/76KOPasyYMXr66aeVkZGht99+Wx9//LF1i31ISIhmzpypJ598UldddZWSkpI0d+5cxcfHKzMz82K+ZwAAoKto6y1pGzZsMJJabFlZWaaysvKcY5LMhg0bjDHGlJWVmdTUVBMZGWnCw8PNNddcY5566ilz6tSpgOf55JNPzI033micTqfp16+fWbRoUYteVqxYYa6++mrjcDjMsGHDzJo1awLGm5ubzdy5c01sbKxxOp1m7NixpqKiotVr5fb5QJ2xZwCA/bTl9TvEGGOCksA6Ab/fr8jISPl8vnb/mGz3lz79+MUPtfrhGzW8X2S7zt1ROmPPAAD7acvrN39rDAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2BZBCAAA2Fabg9DmzZt16623Kj4+XiEhIVq1alXAuDFG8+bNU1xcnCIiIpSenq59+/YF1Hz99deaPHmyXC6XoqKilJ2drePHjwfU7Ny5UzfddJPCw8OVkJCgxYsXt+hl5cqVGjJkiMLDw5WcnKy1a9e2uRcAAGBfbQ5CdXV1SklJ0ZIlS845vnjxYr3wwgtaunSpSktL1aNHD3k8Hp06dcqqmTx5svbs2aOioiKtXr1amzdv1oMPPmiN+/1+jRs3TomJiSorK1NhYaEWLFigV155xarZsmWLJk2apOzsbO3YsUOZmZnKzMzU7t2729QLAACwMXMRJJl3333X2m9ubjZut9sUFhZax2pra43T6TRvvfWWMcaYTz/91EgyH330kVXz17/+1YSEhJgvv/zSGGPMSy+9ZHr37m3q6+utmry8PDN48GBr/8477zQZGRkB/aSmppqHHnqo1b1ciM/nM5KMz+drVX1b7DpcaxLzVptdh2vbfe6O0hl7BgDYT1tev9v1GqHKykp5vV6lp6dbxyIjI5WamqqSkhJJUklJiaKionTDDTdYNenp6QoNDVVpaalVM3r0aDkcDqvG4/GooqJCR48etWrOfp4zNWeepzW9fFN9fb38fn/ABgAAuq52DUJer1eSFBsbG3A8NjbWGvN6vYqJiQkYv+yyyxQdHR1Qc645zn6Ob6s5e/xCvXxTQUGBIiMjrS0hIaEVqwYAAJ0Vd42dJT8/Xz6fz9q++OKLYLcEAAA6ULsGIbfbLUmqrq4OOF5dXW2Nud1u1dTUBIyfPn1aX3/9dUDNueY4+zm+rebs8Qv18k1Op1MulytgAwAAXVe7BqGkpCS53W4VFxdbx/x+v0pLS5WWliZJSktLU21trcrKyqya9evXq7m5WampqVbN5s2b1djYaNUUFRVp8ODB6t27t1Vz9vOcqTnzPK3pBQAA2Fubg9Dx48dVXl6u8vJySf++KLm8vFyHDh1SSEiIZs6cqSeffFJ//vOftWvXLt17772Kj49XZmamJOmaa67R+PHjNXXqVG3btk1///vfNWPGDN11112Kj4+XJN19991yOBzKzs7Wnj17tHz5cj3//PPKzc21+nj00Uf1/vvv6+mnn9bevXu1YMECffzxx5oxY4YktaoXAABgc229JW3Dhg1GUostKyvLGPPv29bnzp1rYmNjjdPpNGPHjjUVFRUBcxw5csRMmjTJ9OzZ07hcLjNlyhRz7NixgJpPPvnE3HjjjcbpdJp+/fqZRYsWtehlxYoV5uqrrzYOh8MMGzbMrFmzJmC8Nb2cD7fPB+qMPQMA7Kctr98hxhgTxBx2SfP7/YqMjJTP52v364V2f+nTj1/8UKsfvlHD+0W269wdpTP2DACwn7a8fnPXGAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsK3Lgt0AOp/9NceD3UKb9O7hUL+oiGC3AQC4BBGE0Gq9ezgU0S1MM5eXB7uVNonoFqZ1s8cQhgAALbR7EBowYIA+//zzFsd/+ctfasmSJbr55pu1adOmgLGHHnpIS5cutfYPHTqk6dOna8OGDerZs6eysrJUUFCgyy77v3Y3btyo3Nxc7dmzRwkJCZozZ47uu+++gHmXLFmiwsJCeb1epaSk6MUXX9SoUaPad8E20i8qQutmj9HRuoZgt9Jq+2uOa+bych2tayAIAQBaaPcg9NFHH6mpqcna3717t/7zP/9TP//5z61jU6dO1RNPPGHtd+/e3fq6qalJGRkZcrvd2rJli6qqqnTvvfeqW7dueuqppyRJlZWVysjI0LRp0/Tmm2+quLhYDzzwgOLi4uTxeCRJy5cvV25urpYuXarU1FQ999xz8ng8qqioUExMTHsv2zb6RUUQKAAAXUa7Xyx9+eWXy+12W9vq1as1aNAgjRkzxqrp3r17QI3L5bLG/vd//1effvqp3njjDV177bWaMGGCfv3rX2vJkiVqaPj3OxFLly5VUlKSnn76aV1zzTWaMWOGfvazn+nZZ5+15nnmmWc0depUTZkyRUOHDtXSpUvVvXt3vfrqq+29ZAAA0El16F1jDQ0NeuONN3T//fcrJCTEOv7mm2+qb9++Gj58uPLz83XixAlrrKSkRMnJyYqNjbWOeTwe+f1+7dmzx6pJT08PeC6Px6OSkhLrecvKygJqQkNDlZ6ebtWcS319vfx+f8AGAAC6rg69WHrVqlWqra0NuHbn7rvvVmJiouLj47Vz507l5eWpoqJC77zzjiTJ6/UGhCBJ1r7X6z1vjd/v18mTJ3X06FE1NTWds2bv3r3f2m9BQYEWLlz4ndcLAAA6lw4NQn/4wx80YcIExcfHW8cefPBB6+vk5GTFxcVp7NixOnDggAYNGtSR7VxQfn6+cnNzrX2/36+EhIQgdgQAADpShwWhzz//XOvWrbPe6fk2qampkqT9+/dr0KBBcrvd2rZtW0BNdXW1JMntdlv/e+bY2TUul0sREREKCwtTWFjYOWvOzHEuTqdTTqezdQsEAACdXoddI/Taa68pJiZGGRkZ560rLy+XJMXFxUmS0tLStGvXLtXU1Fg1RUVFcrlcGjp0qFVTXFwcME9RUZHS0tIkSQ6HQyNHjgyoaW5uVnFxsVUDAADQIUGoublZr732mrKysgJ+98+BAwf061//WmVlZTp48KD+/Oc/695779Xo0aM1YsQISdK4ceM0dOhQ3XPPPfrkk0/0wQcfaM6cOcrJybHerZk2bZr+8Y9/6PHHH9fevXv10ksvacWKFZo1a5b1XLm5ufrd736n119/XZ999pmmT5+uuro6TZkypSOWDAAAOqEO+Whs3bp1OnTokO6///6A4w6HQ+vWrdNzzz2nuro6JSQk6I477tCcOXOsmrCwMK1evVrTp09XWlqaevTooaysrIDfO5SUlKQ1a9Zo1qxZev7553XFFVfo97//vfU7hCRp4sSJ+uc//6l58+bJ6/Xq2muv1fvvv9/iAmoAAGBfIcYYE+wmLlV+v1+RkZHy+XwBv+uoPez+0qcfv/ihVj98o4b3i2zXufF/+D4DgP205fWbvz4PAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABsiyAEAABs67JgNwB8H/bXHA92C23Su4dD/aIigt0GAHR5BCF0ab17OBTRLUwzl5cHu5U2iegWpnWzxxCGAKCDEYTQpfWLitC62WN0tK4h2K202v6a45q5vFxH6xoIQgDQwdo9CC1YsEALFy4MODZ48GDt3btXknTq1CnNnj1bb7/9turr6+XxePTSSy8pNjbWqj906JCmT5+uDRs2qGfPnsrKylJBQYEuu+z/2t24caNyc3O1Z88eJSQkaM6cObrvvvsCnnfJkiUqLCyU1+tVSkqKXnzxRY0aNaq9l4xLXL+oCAIFAOCcOuRi6WHDhqmqqsraPvzwQ2ts1qxZ+stf/qKVK1dq06ZN+uqrr3T77bdb401NTcrIyFBDQ4O2bNmi119/XcuWLdO8efOsmsrKSmVkZOiWW25ReXm5Zs6cqQceeEAffPCBVbN8+XLl5uZq/vz52r59u1JSUuTxeFRTU9MRSwYAAJ2RaWfz5883KSkp5xyrra013bp1MytXrrSOffbZZ0aSKSkpMcYYs3btWhMaGmq8Xq9V8/LLLxuXy2Xq6+uNMcY8/vjjZtiwYQFzT5w40Xg8Hmt/1KhRJicnx9pvamoy8fHxpqCgoNVr8fl8RpLx+Xytfkxr7TpcaxLzVptdh2vbfW50bvxsAMDFacvrd4e8I7Rv3z7Fx8dr4MCBmjx5sg4dOiRJKisrU2Njo9LT063aIUOGqH///iopKZEklZSUKDk5OeCjMo/HI7/frz179lg1Z89xpubMHA0NDSorKwuoCQ0NVXp6ulUDAADQ7tcIpaamatmyZRo8eLCqqqq0cOFC3XTTTdq9e7e8Xq8cDoeioqICHhMbGyuv1ytJ8nq9ASHozPiZsfPV+P1+nTx5UkePHlVTU9M5a85cq3Qu9fX1qq+vt/b9fn/bFg8AADqVdg9CEyZMsL4eMWKEUlNTlZiYqBUrVigi4tK+YLWgoKDFhd4AAKDr6vDfLB0VFaWrr75a+/fvl9vtVkNDg2prawNqqqur5Xa7JUlut1vV1dUtxs+Mna/G5XIpIiJCffv2VVhY2DlrzsxxLvn5+fL5fNb2xRdffKc1AwCAzqHDg9Dx48d14MABxcXFaeTIkerWrZuKi4ut8YqKCh06dEhpaWmSpLS0NO3atSvg7q6ioiK5XC4NHTrUqjl7jjM1Z+ZwOBwaOXJkQE1zc7OKi4utmnNxOp1yuVwBGwAA6LraPQj96le/0qZNm3Tw4EFt2bJFP/3pTxUWFqZJkyYpMjJS2dnZys3N1YYNG1RWVqYpU6YoLS1NP/zhDyVJ48aN09ChQ3XPPffok08+0QcffKA5c+YoJydHTqdTkjRt2jT94x//0OOPP669e/fqpZde0ooVKzRr1iyrj9zcXP3ud7/T66+/rs8++0zTp09XXV2dpkyZ0t5LBgAAnVS7XyN0+PBhTZo0SUeOHNHll1+uG2+8UVu3btXll18uSXr22WcVGhqqO+64I+AXKp4RFham1atXa/r06UpLS1OPHj2UlZWlJ554wqpJSkrSmjVrNGvWLD3//PO64oor9Pvf/14ej8eqmThxov75z39q3rx58nq9uvbaa/X++++3uIAaAADYV4gxxgS7iUuV3+9XZGSkfD5fu39MtvtLn3784oda/fCNGt4vsl3nRufGzwYAXJy2vH53+DVCAAAAlyqCEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsC2CEAAAsK3Lgt0AgHPbX3M82C20Se8eDvWLigh2GwDQJgQh4BLTu4dDEd3CNHN5ebBbaZOIbmFaN3sMYQhAp0IQAi4x/aIitG72GB2tawh2K622v+a4Zi4v19G6BoIQgE6FIARcgvpFRXTKQMHHeQA6G4IQgIvGx3kAOiuCEICLxsd5ADorghCAdtFZP84DYG/8HiEAAGBbBCEAAGBbfDQGwNa40w2wN4IQAFviTjcAEkEIgE1xpxsAiSAEwMa40w0AF0sDAADbIggBAADbIggBAADbIggBAADbavcgVFBQoB/84Afq1auXYmJilJmZqYqKioCam2++WSEhIQHbtGnTAmoOHTqkjIwMde/eXTExMXrsscd0+vTpgJqNGzfq+uuvl9Pp1JVXXqlly5a16GfJkiUaMGCAwsPDlZqaqm3btrX3kgEAQCfV7kFo06ZNysnJ0datW1VUVKTGxkaNGzdOdXV1AXVTp05VVVWVtS1evNgaa2pqUkZGhhoaGrRlyxa9/vrrWrZsmebNm2fVVFZWKiMjQ7fccovKy8s1c+ZMPfDAA/rggw+smuXLlys3N1fz58/X9u3blZKSIo/Ho5qamvZeNgAA6IxMB6upqTGSzKZNm6xjY8aMMY8++ui3Pmbt2rUmNDTUeL1e69jLL79sXC6Xqa+vN8YY8/jjj5thw4YFPG7ixInG4/FY+6NGjTI5OTnWflNTk4mPjzcFBQWt6t3n8xlJxufztaq+LXYdrjWJeavNrsO17T43gK6J/24ArdOW1+8Ov0bI5/NJkqKjowOOv/nmm+rbt6+GDx+u/Px8nThxwhorKSlRcnKyYmNjrWMej0d+v1979uyxatLT0wPm9Hg8KikpkSQ1NDSorKwsoCY0NFTp6elWzTfV19fL7/cHbAAAoOvq0F+o2NzcrJkzZ+pHP/qRhg8fbh2/++67lZiYqPj4eO3cuVN5eXmqqKjQO++8I0nyer0BIUiSte/1es9b4/f7dfLkSR09elRNTU3nrNm7d+85+y0oKNDChQsvbtEAAKDT6NAglJOTo927d+vDDz8MOP7ggw9aXycnJysuLk5jx47VgQMHNGjQoI5s6bzy8/OVm5tr7fv9fiUkJAStHwAA0LE6LAjNmDFDq1ev1ubNm3XFFVectzY1NVWStH//fg0aNEhut7vF3V3V1dWSJLfbbf3vmWNn17hcLkVERCgsLExhYWHnrDkzxzc5nU45nc7WLxIAAHRq7X6NkDFGM2bM0Lvvvqv169crKSnpgo8pLy+XJMXFxUmS0tLStGvXroC7u4qKiuRyuTR06FCrpri4OGCeoqIipaWlSZIcDodGjhwZUNPc3Kzi4mKrBgAA2Fu7vyOUk5OjP/3pT3rvvffUq1cv65qeyMhIRURE6MCBA/rTn/6k//qv/1KfPn20c+dOzZo1S6NHj9aIESMkSePGjdPQoUN1zz33aPHixfJ6vZozZ45ycnKsd2ymTZum3/zmN3r88cd1//33a/369VqxYoXWrFlj9ZKbm6usrCzdcMMNGjVqlJ577jnV1dVpypQp7b1sAADQGbX3LWuSzrm99tprxhhjDh06ZEaPHm2io6ON0+k0V155pXnsscda3OJ28OBBM2HCBBMREWH69u1rZs+ebRobGwNqNmzYYK699lrjcDjMwIEDrec424svvmj69+9vHA6HGTVqlNm6dWur18Lt8wAuJfx3A2idtrx+t/s7QsaY844nJCRo06ZNF5wnMTFRa9euPW/NzTffrB07dpy3ZsaMGZoxY8YFnw8AANgPf2sMAADYFkEIAADYFkEIAADYFkEIAADYVof+ZmkAQPvbX3M82C20We8eDvWLigh2G0ALBCEA6CR693AooluYZi4vD3YrbRbRLUzrZo8hDOGSQxACgE6iX1SE1s0eo6N1DcFupU321xzXzOXlOlrXQBDCJYcgBACdSL+oCMIE0I64WBoAANgWQQgAANgWQQgAANgWQQgAANgWQQgAANgWQQgAANgWQQgAANgWQQgAANgWQQgAANgWQQgAANgWQQgAANgWQQgAANgWQQgAANgWQQgAANgWQQgAANgWQQgAANjWZcFuAABgD/trjge7hTbp3cOhflERwW4DHYwgBADoUL17OBTRLUwzl5cHu5U2iegWpnWzxxCGujiCEACgQ/WLitC62WN0tK4h2K202v6a45q5vFxH6xoIQl0cQQgA0OH6RUUQKHBJ4mJpAABgWwQhAABgWwQhAABgW7YIQkuWLNGAAQMUHh6u1NRUbdu2LdgtAQCAS0CXD0LLly9Xbm6u5s+fr+3btyslJUUej0c1NTXBbg0AAARZl79r7JlnntHUqVM1ZcoUSdLSpUu1Zs0avfrqq/rv//7vIHcHALiU8Usgu74uHYQaGhpUVlam/Px861hoaKjS09NVUlLSor6+vl719fXWvs/nkyT5/f527+34Mb+a60/o+DG//P6Qdp8fAPDdXdZ0So7mU3rkj1uC3UqbhHcL1XN3Xafo7t2C3UqrXd7Tqctd4e0655nXbWPMBWu7dBD617/+paamJsXGxgYcj42N1d69e1vUFxQUaOHChS2OJyQkdFiPac912NQAABvKKAx2B5eOY8eOKTIy8rw1XToItVV+fr5yc3Ot/ebmZn399dfq06ePQkLa510bv9+vhIQEffHFF3K5XO0y56XKTmuV7LVeO61VYr1dmZ3WKtlnvcYYHTt2TPHx8Res7dJBqG/fvgoLC1N1dXXA8erqarnd7hb1TqdTTqcz4FhUVFSH9OZyubr0D+HZ7LRWyV7rtdNaJdbbldlprZI91nuhd4LO6NJ3jTkcDo0cOVLFxcXWsebmZhUXFystLS2InQEAgEtBl35HSJJyc3OVlZWlG264QaNGjdJzzz2nuro66y4yAABgX10+CE2cOFH//Oc/NW/ePHm9Xl177bV6//33W1xA/X1xOp2aP39+i4/guiI7rVWy13rttFaJ9XZldlqrZL/1tkaIac29ZQAAAF1Ql75GCAAA4HwIQgAAwLYIQgAAwLYIQgAAwLYIQt+jJUuWaMCAAQoPD1dqaqq2bdsW7JYuWkFBgX7wgx+oV69eiomJUWZmpioqKgJqbr75ZoWEhARs06ZNC1LHF2fBggUt1jJkyBBr/NSpU8rJyVGfPn3Us2dP3XHHHS1+oWdnMmDAgBbrDQkJUU5OjqTOfW43b96sW2+9VfHx8QoJCdGqVasCxo0xmjdvnuLi4hQREaH09HTt27cvoObrr7/W5MmT5XK5FBUVpezsbB0/fmn+kc7zrbexsVF5eXlKTk5Wjx49FB8fr3vvvVdfffVVwBzn+nlYtGjR97yS1rnQ+b3vvvtarGX8+PEBNZ3l/F5oref6NxwSEqLCwv/7Wxyd6dy2N4LQ92T58uXKzc3V/PnztX37dqWkpMjj8aimpibYrV2UTZs2KScnR1u3blVRUZEaGxs1btw41dXVBdRNnTpVVVVV1rZ48eIgdXzxhg0bFrCWDz/80BqbNWuW/vKXv2jlypXatGmTvvrqK91+++1B7PbifPTRRwFrLSoqkiT9/Oc/t2o667mtq6tTSkqKlixZcs7xxYsX64UXXtDSpUtVWlqqHj16yOPx6NSpU1bN5MmTtWfPHhUVFWn16tXavHmzHnzwwe9rCW1yvvWeOHFC27dv19y5c7V9+3a98847qqio0E9+8pMWtU888UTA+X744Ye/j/bb7ELnV5LGjx8fsJa33norYLyznN8LrfXsNVZVVenVV19VSEiI7rjjjoC6znJu253B92LUqFEmJyfH2m9qajLx8fGmoKAgiF21v5qaGiPJbNq0yTo2ZswY8+ijjwavqXY0f/58k5KScs6x2tpa061bN7Ny5Urr2GeffWYkmZKSku+pw4716KOPmkGDBpnm5mZjTNc5t5LMu+++a+03Nzcbt9ttCgsLrWO1tbXG6XSat956yxhjzKeffmokmY8++siq+etf/2pCQkLMl19++b31/l18c73nsm3bNiPJfP7559axxMRE8+yzz3Zscx3gXOvNysoyt91227c+prOe39ac29tuu838x3/8R8Cxznpu2wPvCH0PGhoaVFZWpvT0dOtYaGio0tPTVVJSEsTO2p/P55MkRUdHBxx/88031bdvXw0fPlz5+fk6ceJEMNprF/v27VN8fLwGDhyoyZMn69ChQ5KksrIyNTY2BpznIUOGqH///l3iPDc0NOiNN97Q/fffH/BHiLvSuT2jsrJSXq834FxGRkYqNTXVOpclJSWKiorSDTfcYNWkp6crNDRUpaWl33vP7c3n8ykkJKTF31tctGiR+vTpo+uuu06FhYU6ffp0cBpsBxs3blRMTIwGDx6s6dOn68iRI9ZYVz2/1dXVWrNmjbKzs1uMdaVz2xZd/jdLXwr+9a9/qampqcVvs46NjdXevXuD1FX7a25u1syZM/WjH/1Iw4cPt47ffffdSkxMVHx8vHbu3Km8vDxVVFTonXfeCWK3301qaqqWLVumwYMHq6qqSgsXLtRNN92k3bt3y+v1yuFwtHjhiI2NldfrDU7D7WjVqlWqra3VfffdZx3rSuf2bGfO17n+zZ4Z83q9iomJCRi/7LLLFB0d3enP96lTp5SXl6dJkyYF/GHORx55RNdff72io6O1ZcsW5efnq6qqSs8880wQu/1uxo8fr9tvv11JSUk6cOCA/t//+3+aMGGCSkpKFBYW1mXP7+uvv65evXq1+Mi+K53btiIIod3k5ORo9+7dAdfMSAr4TD05OVlxcXEaO3asDhw4oEGDBn3fbV6UCRMmWF+PGDFCqampSkxM1IoVKxQRERHEzjreH/7wB02YMEHx8fHWsa50bvFvjY2NuvPOO2WM0csvvxwwlpuba309YsQIORwOPfTQQyooKOh0f7Lhrrvusr5OTk7WiBEjNGjQIG3cuFFjx44NYmcd69VXX9XkyZMVHh4ecLwrndu24qOx70Hfvn0VFhbW4u6h6upqud3uIHXVvmbMmKHVq1drw4YNuuKKK85bm5qaKknav3//99Fah4qKitLVV1+t/fv3y+12q6GhQbW1tQE1XeE8f/7551q3bp0eeOCB89Z1lXN75nyd79+s2+1ucbPD6dOn9fXXX3fa830mBH3++ecqKioKeDfoXFJTU3X69GkdPHjw+2mwAw0cOFB9+/a1fna74vn929/+poqKigv+O5a61rm9EILQ98DhcGjkyJEqLi62jjU3N6u4uFhpaWlB7OziGWM0Y8YMvfvuu1q/fr2SkpIu+Jjy8nJJUlxcXAd31/GOHz+uAwcOKC4uTiNHjlS3bt0CznNFRYUOHTrU6c/za6+9ppiYGGVkZJy3rquc26SkJLnd7oBz6ff7VVpaap3LtLQ01dbWqqyszKpZv369mpubrUDYmZwJQfv27dO6devUp0+fCz6mvLxcoaGhLT5C6owOHz6sI0eOWD+7Xe38Sv9+V3fkyJFKSUm5YG1XOrcXFOyrte3i7bffNk6n0yxbtsx8+umn5sEHHzRRUVHG6/UGu7WLMn36dBMZGWk2btxoqqqqrO3EiRPGGGP2799vnnjiCfPxxx+byspK895775mBAwea0aNHB7nz72b27Nlm48aNprKy0vz973836enppm/fvqampsYYY8y0adNM//79zfr1683HH39s0tLSTFpaWpC7vjhNTU2mf//+Ji8vL+B4Zz+3x44dMzt27DA7duwwkswzzzxjduzYYd0ltWjRIhMVFWXee+89s3PnTnPbbbeZpKQkc/LkSWuO8ePHm+uuu86UlpaaDz/80Fx11VVm0qRJwVrSeZ1vvQ0NDeYnP/mJueKKK0x5eXnAv+X6+npjjDFbtmwxzz77rCkvLzcHDhwwb7zxhrn88svNvffeG+SVndv51nvs2DHzq1/9ypSUlJjKykqzbt06c/3115urrrrKnDp1ypqjs5zfC/0sG2OMz+cz3bt3Ny+//HKLx3e2c9veCELfoxdffNH079/fOBwOM2rUKLN169Zgt3TRJJ1ze+2114wxxhw6dMiMHj3aREdHG6fTaa688krz2GOPGZ/PF9zGv6OJEyeauLg443A4TL9+/czEiRPN/v37rfGTJ0+aX/7yl6Z3796me/fu5qc//ampqqoKYscX74MPPjCSTEVFRcDxzn5uN2zYcM6f3aysLGPMv2+hnzt3romNjTVOp9OMHTu2xffgyJEjZtKkSaZnz57G5XKZKVOmmGPHjgVhNRd2vvVWVlZ+67/lDRs2GGOMKSsrM6mpqSYyMtKEh4eba665xjz11FMBweFScr71njhxwowbN85cfvnlplu3biYxMdFMnTq1xf8x7Szn90I/y8YY89vf/tZERESY2traFo/vbOe2vYUYY0yHvuUEAABwieIaIQAAYFsEIQAAYFsEIQAAYFsEIQAAYFsEIQAAYFsEIQAAYFsEIQAAYFsEIQAAYFsEIQAAYFsEIQAAYFsEIQAAYFsEIQAAYFv/H1nBbmdzi309AAAAAElFTkSuQmCC\n"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "pixel_counts = train.groupby(['lat', 'lon']).size()\n",
    "counts, bins = np.histogram(pixel_counts[0])\n",
    "plt.stairs(counts, bins)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "source": [],
   "metadata": {
    "id": "zJL1npdxHtp4",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": 22,
   "outputs": []
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "outputs": [],
   "source": [],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "source": [
    "X = np.empty([0, n_input, n_features])\n",
    "y = np.empty([0, 2])\n",
    "\n",
    "print(X, y)\n",
    "\n",
    "counter = 0\n",
    "savept = 0\n",
    "savept = 0\n",
    "print(time.strftime(\"%Y-%m-%d %H:%M:%S\", time.localtime()))"
   ],
   "metadata": {
    "id": "PVy4KpjrU1LY",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": 40,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[] []\n",
      "2022-12-20 12:17:25\n"
     ]
    }
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "outputs": [],
   "source": [],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Length of df of unique latlon paris that meet cutoff:  85603\n",
      "2022-12-20 18:05:19\n",
      "Save last: \n"
     ]
    },
    {
     "ename": "TypeError",
     "evalue": "unsupported operand type(s) for -: 'str' and 'str'",
     "output_type": "error",
     "traceback": [
      "\u001B[1;31m---------------------------------------------------------------------------\u001B[0m",
      "\u001B[1;31mTypeError\u001B[0m                                 Traceback (most recent call last)",
      "Cell \u001B[1;32mIn[120], line 13\u001B[0m\n\u001B[0;32m     10\u001B[0m savept \u001B[38;5;241m=\u001B[39m \u001B[38;5;241m85\u001B[39m\n\u001B[0;32m     12\u001B[0m \u001B[38;5;28mprint\u001B[39m(time\u001B[38;5;241m.\u001B[39mstrftime(\u001B[38;5;124m\"\u001B[39m\u001B[38;5;124m%\u001B[39m\u001B[38;5;124mY-\u001B[39m\u001B[38;5;124m%\u001B[39m\u001B[38;5;124mm-\u001B[39m\u001B[38;5;132;01m%d\u001B[39;00m\u001B[38;5;124m \u001B[39m\u001B[38;5;124m%\u001B[39m\u001B[38;5;124mH:\u001B[39m\u001B[38;5;124m%\u001B[39m\u001B[38;5;124mM:\u001B[39m\u001B[38;5;124m%\u001B[39m\u001B[38;5;124mS\u001B[39m\u001B[38;5;124m\"\u001B[39m, time\u001B[38;5;241m.\u001B[39mlocaltime()))\n\u001B[1;32m---> 13\u001B[0m \u001B[43mdata_prep\u001B[49m\u001B[43m(\u001B[49m\u001B[43mtrain\u001B[49m\u001B[43m,\u001B[49m\u001B[43m \u001B[49m\u001B[43mlatlon\u001B[49m\u001B[43m,\u001B[49m\u001B[43m \u001B[49m\u001B[43mpath\u001B[49m\u001B[43m,\u001B[49m\u001B[43m \u001B[49m\u001B[43mn_input\u001B[49m\u001B[43m,\u001B[49m\u001B[43m \u001B[49m\u001B[43mn_features\u001B[49m\u001B[43m,\u001B[49m\u001B[43m \u001B[49m\u001B[43mX\u001B[49m\u001B[43m,\u001B[49m\u001B[43m \u001B[49m\u001B[43my\u001B[49m\u001B[43m,\u001B[49m\u001B[43m \u001B[49m\u001B[43mcounter\u001B[49m\u001B[43m,\u001B[49m\u001B[43m \u001B[49m\u001B[43msavept\u001B[49m\u001B[43m)\u001B[49m\n",
      "Cell \u001B[1;32mIn[119], line 24\u001B[0m, in \u001B[0;36mdata_prep\u001B[1;34m(bigdf, latlon, path, n_input, n_features, X, y, counter, savept)\u001B[0m\n\u001B[0;32m     22\u001B[0m       np\u001B[38;5;241m.\u001B[39msave(f, X)\n\u001B[0;32m     23\u001B[0m       np\u001B[38;5;241m.\u001B[39msave(f, y)\n\u001B[1;32m---> 24\u001B[0m \u001B[38;5;28mprint\u001B[39m(\u001B[38;5;124m\"\u001B[39m\u001B[38;5;124mDone! And it only took: \u001B[39m\u001B[38;5;124m\"\u001B[39m, \u001B[43mtime\u001B[49m\u001B[38;5;241;43m.\u001B[39;49m\u001B[43mstrftime\u001B[49m\u001B[43m(\u001B[49m\u001B[38;5;124;43m\"\u001B[39;49m\u001B[38;5;124;43m%\u001B[39;49m\u001B[38;5;124;43mY-\u001B[39;49m\u001B[38;5;124;43m%\u001B[39;49m\u001B[38;5;124;43mm-\u001B[39;49m\u001B[38;5;132;43;01m%d\u001B[39;49;00m\u001B[38;5;124;43m \u001B[39;49m\u001B[38;5;124;43m%\u001B[39;49m\u001B[38;5;124;43mH:\u001B[39;49m\u001B[38;5;124;43m%\u001B[39;49m\u001B[38;5;124;43mM:\u001B[39;49m\u001B[38;5;124;43m%\u001B[39;49m\u001B[38;5;124;43mS\u001B[39;49m\u001B[38;5;124;43m\"\u001B[39;49m\u001B[43m,\u001B[49m\u001B[43m \u001B[49m\u001B[43mtime\u001B[49m\u001B[38;5;241;43m.\u001B[39;49m\u001B[43mlocaltime\u001B[49m\u001B[43m(\u001B[49m\u001B[43m)\u001B[49m\u001B[43m)\u001B[49m\u001B[43m \u001B[49m\u001B[38;5;241;43m-\u001B[39;49m\u001B[43m \u001B[49m\u001B[43mstart\u001B[49m)\n",
      "\u001B[1;31mTypeError\u001B[0m: unsupported operand type(s) for -: 'str' and 'str'"
     ]
    }
   ],
   "source": [
    "n_input = 80\n",
    "n_features = 7\n",
    "\n",
    "path = 'LSTMData80Sparse/XY_arrays_LSTM_80Sparse_'\n",
    "\n",
    "latlon = latlon_setup(train, n_input)\n",
    "X = np.empty([0, n_input, n_features])\n",
    "y = np.empty([0, 2])\n",
    "counter = 85000\n",
    "savept = 85\n",
    "\n",
    "print(time.strftime(\"%Y-%m-%d %H:%M:%S\", time.localtime()))\n",
    "data_prep(train, latlon, path, n_input, n_features, X, y, counter, savept)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "source": [
    "# n_input = 60\n",
    "# n_features = 7\n",
    "#\n",
    "# path = 'LSTMData60Sparse/XY_arrays_LSTM_60Sparse_'\n",
    "#\n",
    "# latlon = latlon_setup(train, n_input)\n",
    "# X = np.empty([0, n_input, n_features])\n",
    "# y = np.empty([0, 2])\n",
    "# counter = 0\n",
    "# savept = 0\n",
    "#\n",
    "# print(time.strftime(\"%Y-%m-%d %H:%M:%S\", time.localtime()))\n",
    "# data_prep(train, latlon, path, n_input, n_features, X, y, counter, savept)"
   ],
   "metadata": {
    "id": "wkQWZIGLPlUo",
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "outputId": "98d0520c-74be-4e5a-8071-ef72febc768e",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": 41,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Counter:  10000 , Shape X:  (40763, 60, 7) 2022-12-20 12:25:30\n",
      "Wipeout:  0\n",
      "Counter:  20000 , Shape X:  (40738, 60, 7) 2022-12-20 12:33:48\n",
      "Wipeout:  1\n",
      "Counter:  30000 , Shape X:  (40793, 60, 7) 2022-12-20 12:42:34\n",
      "Wipeout:  2\n",
      "Counter:  40000 , Shape X:  (40669, 60, 7) 2022-12-20 12:51:36\n",
      "Wipeout:  3\n",
      "Counter:  50000 , Shape X:  (40449, 60, 7) 2022-12-20 13:00:42\n",
      "Wipeout:  4\n",
      "Counter:  60000 , Shape X:  (40742, 60, 7) 2022-12-20 13:09:49\n",
      "Wipeout:  5\n",
      "Counter:  70000 , Shape X:  (40791, 60, 7) 2022-12-20 13:18:56\n",
      "Wipeout:  6\n",
      "Counter:  80000 , Shape X:  (40857, 60, 7) 2022-12-20 13:27:36\n",
      "Wipeout:  7\n",
      "Counter:  90000 , Shape X:  (40889, 60, 7) 2022-12-20 13:36:13\n",
      "Wipeout:  8\n",
      "Counter:  100000 , Shape X:  (40264, 60, 7) 2022-12-20 13:45:03\n",
      "Wipeout:  9\n",
      "Counter:  110000 , Shape X:  (40733, 60, 7) 2022-12-20 13:54:12\n",
      "Wipeout:  10\n",
      "Counter:  120000 , Shape X:  (41158, 60, 7) 2022-12-20 14:02:55\n",
      "Wipeout:  11\n"
     ]
    }
   ]
  },
  {
   "cell_type": "code",
   "source": [
    "print(counter)"
   ],
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "26HhGunxWeKj",
    "outputId": "e151d6ad-eea6-4786-d07c-270620f317b1",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": 42,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "123873\n"
     ]
    }
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "outputs": [],
   "source": [
    "X_train = np.empty([0, n_input, n_features])\n",
    "y_train = np.empty([0,2])\n",
    "X_val = np.empty([0, n_input, n_features])\n",
    "y_val = np.empty([0,2])"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_0.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_0.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_1.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_1.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_10.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_10.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_11.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_11.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_12.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_12.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_13.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_13.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_14.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_14.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_15.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_15.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_16.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_16.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_17.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_17.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_18.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_18.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_19.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_19.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_2.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_2.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_20.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_20.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_21.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_21.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_22.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_22.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_23.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_23.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_24.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_24.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_25.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_25.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_26.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_26.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_27.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_27.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_28.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_28.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_29.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_29.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_3.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_3.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_30.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_30.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_31.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_31.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_32.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_32.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_33.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_33.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_34.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_34.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_35.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_35.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_36.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_36.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_37.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_37.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_38.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_38.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_39.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_39.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_4.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_4.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_40.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_40.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_41.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_41.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_42.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_42.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_43.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_43.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_44.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_44.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_45.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_45.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_46.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_46.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_47.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_47.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_48.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_48.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_49.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_49.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_5.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_5.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_50.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_50.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_51.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_51.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_52.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_52.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_53.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_53.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_54.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_54.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_55.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_55.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_56.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_56.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_57.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_57.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_58.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_58.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_59.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_59.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_6.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_6.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_60.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_60.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_61.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_61.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_62.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_62.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_63.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_63.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_64.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_64.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_65.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_65.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_7.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_7.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_8.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_8.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_80Sparse_LSTM_Train_9.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_80Sparse_LSTM_Train_9.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_LSTM_80Sparse_66.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_LSTM_80Sparse_66.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_LSTM_80Sparse_67.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_LSTM_80Sparse_67.npy'>\n",
      "LSTMData80Sparse\\XY_arrays_LSTM_80Sparse_68.npy\n",
      "Train:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_LSTM_80Sparse_68.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_LSTM_80Sparse_69.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_LSTM_80Sparse_70.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_LSTM_80Sparse_71.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_LSTM_80Sparse_72.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_LSTM_80Sparse_73.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_LSTM_80Sparse_74.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_LSTM_80Sparse_75.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_LSTM_80Sparse_76.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_LSTM_80Sparse_77.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_LSTM_80Sparse_78.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_LSTM_80Sparse_79.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_LSTM_80Sparse_80.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_LSTM_80Sparse_81.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_LSTM_80Sparse_82.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_LSTM_80Sparse_83.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_LSTM_80Sparse_84.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTMData80Sparse\\\\XY_arrays_LSTM_80Sparse__last.npy'>\n"
     ]
    }
   ],
   "source": [
    "path = 'LSTMData80Sparse'\n",
    "filecount = len(os.listdir(path))\n",
    "for filename in os.listdir(path):\n",
    "    filecount = filecount-1\n",
    "    f = os.path.join(path, filename)\n",
    "    if filecount > 16:\n",
    "        print(f)\n",
    "        with open(f, 'rb') as f: #Saves two arrays in same file, arrays can be opened separately or together.\n",
    "            X_load = np.load(f)\n",
    "            y_load = np.load(f)\n",
    "        X_train = np.vstack([X_train, X_load]) #Append your arrays\n",
    "        y_train = np.concatenate([y_train, y_load])\n",
    "        print(\"Train: \", f)\n",
    "    else:\n",
    "        with open(f, 'rb') as f: #Saves two arrays in same file, arrays can be opened separately or together.\n",
    "            X_load = np.load(f)\n",
    "            y_load = np.load(f)\n",
    "        X_val = np.vstack([X_val, X_load]) #Append your arrays\n",
    "        y_val = np.concatenate([y_val, y_load])\n",
    "        print(\"Val: \", f)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(220426, 80, 7)\n",
      "(220426, 2)\n",
      "(53072, 80, 7)\n",
      "(53072, 2)\n",
      "0.19404895099781352\n"
     ]
    }
   ],
   "source": [
    "print(X_train.shape) #(4352938, 60, 7)\n",
    "print(y_train.shape)\n",
    "print(X_val.shape)\n",
    "print(y_val.shape)\n",
    "print(X_val.shape[0] / (X_val.shape[0] + X_train.shape[0])) #60: 0.192, 80: 0.194\n",
    "with open((path + '/XY_80Sparse_Full_XYTrainVal.npy'), 'wb') as f:\n",
    "      np.save(f, X_train)\n",
    "      np.save(f, y_train)\n",
    "      np.save(f, X_val)\n",
    "      np.save(f, y_val)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Training - 60 Sparse"
   ],
   "metadata": {
    "id": "33keUIUgLdS8",
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "source": [
    "#https://towardsdatascience.com/choosing-the-right-hyperparameters-for-a-simple-lstm-using-keras-f8e9ed76f046\n",
    "hidden_nodes_ex = int(2/3 * (n_input * n_features * y_train.shape[1]))\n",
    "hidden_nodes_ex2 = int(X_train.shape[0] / 10 * (n_input * n_features * y_train.shape[1]))\n",
    "print(f\"One approximation of the number of hidden nodes is {hidden_nodes_ex}.\")\n",
    "print(f\"Another approximation of the number of hidden nodes is {hidden_nodes_ex2}.\")\n",
    "\n",
    "hidden_nodes = 60\n",
    "print(f\"We will be using {hidden_nodes}, as this is reasonable to train and the same number we used for the DNN.\")"
   ],
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "p-e7XTPjW4TS",
    "outputId": "464984e2-b1ed-4521-f314-d4cbccc94a46",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": 53,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "One approximation of the number of hidden nodes is 560.\n",
      "Another approximation of the number of hidden nodes is 34246212.\n",
      "We will be using 60, as this is reasonable to train and the same number we used for the DNN.\n"
     ]
    }
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(407693, 60, 7)\n",
      "(407693, 2)\n",
      "(97028, 60, 7)\n",
      "(97028, 2)\n"
     ]
    }
   ],
   "source": [
    "print(X_train.shape) #(4352938, 60, 7), (3508883, 60, 7), (407693, 60, 7)\n",
    "print(y_train.shape)\n",
    "print(X_val.shape)\n",
    "print(y_val.shape)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "source": [],
   "metadata": {
    "id": "cCkZOpWwLc7r",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": 55,
   "outputs": []
  },
  {
   "cell_type": "code",
   "source": [
    "model = model_lstm()\n",
    "model.summary()"
   ],
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "XPJkCUzFezo0",
    "outputId": "b31cf317-8a69-4354-cf88-b67773d8c1ce",
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "execution_count": 56,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " lstm_2 (LSTM)               (None, 60, 60)            16320     \n",
      "                                                                 \n",
      " dropout_2 (Dropout)         (None, 60, 60)            0         \n",
      "                                                                 \n",
      " lstm_3 (LSTM)               (None, 60)                29040     \n",
      "                                                                 \n",
      " dropout_3 (Dropout)         (None, 60)                0         \n",
      "                                                                 \n",
      " dense_1 (Dense)             (None, 2)                 122       \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 45,482\n",
      "Trainable params: 45,482\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "outputs": [],
   "source": [
    "batch_size = 500\n",
    "\n",
    "callbacks_list = [\n",
    "    tf.keras.callbacks.EarlyStopping(\n",
    "        monitor=\"val_loss\",\n",
    "        patience=3\n",
    "    ),\n",
    "    tf.keras.callbacks.ModelCheckpoint(\n",
    "        filepath=\"ModelCheckpoints60/model.{epoch:02d}-{val_loss:.2f}.h5\",\n",
    "        monitor=\"val_loss\",\n",
    "        mode=\"min\",\n",
    "        save_best_only=True,\n",
    "    )\n",
    "]"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Time:  14:14:29\n",
      "Epoch 1/20\n",
      "816/816 [==============================] - 461s 563ms/step - loss: 0.0250 - val_loss: 0.0195\n",
      "Epoch 2/20\n",
      "816/816 [==============================] - 407s 499ms/step - loss: 0.0190 - val_loss: 0.0179\n",
      "Epoch 3/20\n",
      "816/816 [==============================] - 369s 452ms/step - loss: 0.0168 - val_loss: 0.0151\n",
      "Epoch 4/20\n",
      "816/816 [==============================] - 351s 430ms/step - loss: 0.0159 - val_loss: 0.0149\n",
      "Epoch 5/20\n",
      "816/816 [==============================] - 327s 401ms/step - loss: 0.0150 - val_loss: 0.0142\n",
      "Epoch 6/20\n",
      "816/816 [==============================] - 346s 424ms/step - loss: 0.0144 - val_loss: 0.0133\n",
      "Epoch 7/20\n",
      "816/816 [==============================] - 347s 425ms/step - loss: 0.0140 - val_loss: 0.0134\n",
      "Epoch 8/20\n",
      "816/816 [==============================] - 355s 435ms/step - loss: 0.0138 - val_loss: 0.0139\n",
      "Epoch 9/20\n",
      "816/816 [==============================] - 337s 412ms/step - loss: 0.0136 - val_loss: 0.0150\n",
      "Time:  15:09:29\n"
     ]
    }
   ],
   "source": [
    "print(\"Time: \", time.strftime(\"%H:%M:%S\", time.localtime()))\n",
    "\n",
    "history_retrieval = model.fit(X_train, y_train, epochs=20, validation_data=(X_val, y_val),\n",
    "                               batch_size = batch_size, callbacks=callbacks_list, verbose=1)\n",
    "\n",
    "print(\"Time: \", time.strftime(\"%H:%M:%S\", time.localtime()))"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "outputs": [
    {
     "data": {
      "text/plain": "[<matplotlib.lines.Line2D at 0x23f613a5df0>]"
     },
     "execution_count": 59,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": "<Figure size 640x480 with 1 Axes>",
      "image/png": "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\n"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss_per_epoch = model.history.history['loss']\n",
    "plt.plot(range(len(loss_per_epoch)),loss_per_epoch)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "outputs": [],
   "source": [
    "model.save('ModelCheckpoints60/model60sparse_final.h5')"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "816/816 [==============================] - 71s 88ms/step - loss: 0.0152\n"
     ]
    },
    {
     "data": {
      "text/plain": "0.015179895795881748"
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(X_train, y_train, batch_size=batch_size, callbacks=callbacks_list, verbose=1)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "195/195 [==============================] - 16s 80ms/step - loss: 0.0150\n"
     ]
    },
    {
     "data": {
      "text/plain": "0.015012651681900024"
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(X_val, y_val, batch_size=batch_size, callbacks=callbacks_list, verbose=1)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_0.npy'> (3491, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_1.npy'> (6941, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_10.npy'> (10496, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_11.npy'> (13821, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_12.npy'> (17316, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_13.npy'> (20822, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_14.npy'> (24251, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_15.npy'> (27704, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_16.npy'> (31017, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_17.npy'> (34436, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_18.npy'> (37725, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_19.npy'> (41288, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_2.npy'> (44719, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_20.npy'> (48156, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_21.npy'> (51638, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_22.npy'> (55112, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_23.npy'> (58585, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_24.npy'> (62011, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_25.npy'> (65514, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_26.npy'> (68926, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_27.npy'> (72322, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_28.npy'> (75792, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_29.npy'> (79249, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_3.npy'> (82705, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_30.npy'> (86160, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_31.npy'> (89664, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_32.npy'> (93192, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_33.npy'> (96691, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_34.npy'> (100043, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_35.npy'> (103533, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_36.npy'> (106977, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_37.npy'> (110316, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_38.npy'> (113807, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_39.npy'> (117257, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_4.npy'> (120633, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_40.npy'> (124064, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_41.npy'> (127520, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_42.npy'> (130896, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_43.npy'> (134395, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_44.npy'> (137719, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_45.npy'> (141202, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_46.npy'> (144688, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_47.npy'> (148030, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_48.npy'> (151585, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_49.npy'> (154910, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_5.npy'> (158409, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_50.npy'> (161904, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_51.npy'> (165410, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_52.npy'> (168839, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_53.npy'> (172292, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_54.npy'> (175605, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_55.npy'> (179024, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_56.npy'> (182313, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_57.npy'> (185876, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_58.npy'> (189313, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_59.npy'> (192795, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_6.npy'> (196119, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_60.npy'> (199593, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_61.npy'> (203066, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_62.npy'> (206492, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_63.npy'> (209995, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_64.npy'> (213407, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_65.npy'> (216803, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_66.npy'> (220273, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_67.npy'> (223730, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_68.npy'> (227185, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_69.npy'> (230689, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_7.npy'> (234172, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_70.npy'> (237700, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_71.npy'> (241199, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_72.npy'> (244551, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_73.npy'> (248041, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_74.npy'> (251485, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_75.npy'> (254824, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_76.npy'> (258221, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_77.npy'> (261669, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_78.npy'> (265141, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_79.npy'> (268520, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_8.npy'> (272006, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_80.npy'> (275490, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_81.npy'> (278998, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_82.npy'> (282360, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_83.npy'> (285764, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_84.npy'> (289213, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_85.npy'> (292550, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_86.npy'> (295980, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_87.npy'> (299334, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_9.npy'> (302676, 60, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_60Sparse_Test\\\\XY_arrays_60Sparse_LSTM_Test_last.npy'> (302794, 60, 7)\n"
     ]
    }
   ],
   "source": [
    "#Compile test data:\n",
    "X_test = np.empty([0, n_input, n_features])\n",
    "y_test = np.empty([0,2])\n",
    "path = 'LSTM_Data_60Sparse_Test'\n",
    "filecount = len(os.listdir(path))\n",
    "for filename in os.listdir(path):\n",
    "    filecount = filecount-1\n",
    "    p = os.path.join(path, filename)\n",
    "    with open(p, 'rb') as f: #Saves two arrays in same file, arrays can be opened separately or together.\n",
    "        X_load = np.load(f)\n",
    "        y_load = np.load(f)\n",
    "    X_test = np.vstack([X_test, X_load]) #Append your arrays\n",
    "    y_test = np.concatenate([y_test, y_load])\n",
    "    print(\"Test: \", f, X_test.shape)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(302794, 60, 7)\n",
      "(302794, 2)\n"
     ]
    }
   ],
   "source": [
    "print(X_test.shape)\n",
    "print(y_test.shape)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "606/606 [==============================] - 45s 73ms/step\n"
     ]
    }
   ],
   "source": [
    "y_pred = model.predict(X_test, batch_size=batch_size, callbacks=callbacks_list, verbose=1)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(302794, 2)\n",
      "[[0.23307545 0.4318809 ]\n",
      " [0.2775002  0.4064901 ]\n",
      " [0.15507063 0.21285887]\n",
      " [0.32374465 0.4366291 ]\n",
      " [0.41131687 0.22793666]]\n",
      "[[0.28613892 0.22324183]\n",
      " [0.30487439 0.22310174]\n",
      " [0.14736087 0.22358704]\n",
      " [0.30502508 0.22355998]\n",
      " [0.3660039  0.22322477]]\n"
     ]
    }
   ],
   "source": [
    "print(y_pred.shape)\n",
    "print(y_pred[0:5,])\n",
    "print(y_test[0:5]) #is Density then VOD:"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Snow Density Performance: \n",
      "MAE:  0.06776746113589963\n",
      "MSE:  0.009728605803439918\n",
      "RMSE:  0.09863369507141015\n",
      "R2:  0.5597049632701769\n"
     ]
    }
   ],
   "source": [
    "#Snow density:\n",
    "mae = skm.mean_absolute_error(y_test[:,0], y_pred[:,0])\n",
    "mse = skm.mean_squared_error(y_test[:,0], y_pred[:,0])\n",
    "rmse = skm.mean_squared_error(y_test[:,0], y_pred[:,0], squared = False)\n",
    "r2 = skm.r2_score(y_test[:,0], y_pred[:,0])\n",
    "print(\"Snow Density Performance: \")\n",
    "print(\"MAE: \", mae)\n",
    "print(\"MSE: \", mse)\n",
    "print(\"RMSE: \", rmse)\n",
    "print(\"R2: \", r2)\n"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "VOD Performance: \n",
      "MAE:  0.08803249813027872\n",
      "MSE:  0.025105152039051885\n",
      "RMSE:  0.1584460540343365\n",
      "R2:  0.30884914547654385\n"
     ]
    }
   ],
   "source": [
    "#Snow density:\n",
    "mae2 = skm.mean_absolute_error(y_test[:,1], y_pred[:,1])\n",
    "mse2 = skm.mean_squared_error(y_test[:,1], y_pred[:,1])\n",
    "rmse2 = skm.mean_squared_error(y_test[:,1], y_pred[:,1], squared = False)\n",
    "r22 = skm.r2_score(y_test[:,1], y_pred[:,1])\n",
    "print(\"VOD Performance: \")\n",
    "print(\"MAE: \", mae2)\n",
    "print(\"MSE: \", mse2)\n",
    "print(\"RMSE: \", rmse2)\n",
    "print(\"R2: \", r22)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Unscaling results"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "outputs": [
    {
     "data": {
      "text/plain": "array([[0.16662979, 0.24016423],\n       [0.17686891, 0.22891788],\n       [0.18052546, 0.2422458 ],\n       ...,\n       [0.25501974, 0.08590267],\n       [0.2623392 , 0.0891739 ],\n       [0.26668053, 0.0891739 ]])"
     },
     "execution_count": 100,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#DataPrep: Scaling\n",
    "data = pd.read_parquet('processed_data_with_datetime.parquet')\n",
    "scaler = MinMaxScaler()\n",
    "data_scaled = data[['density', 'voc']] #We want lat, lon and date unscaled, as these will be used for sorting, not training.\n",
    "data_scaled = scaler.fit_transform(data_scaled)\n",
    "data_scaled"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "outputs": [
    {
     "data": {
      "text/plain": "array([[2.00148606e+02, 3.34834448e-01],\n       [2.06706021e+02, 3.34624320e-01],\n       [1.51576290e+02, 3.35352222e-01],\n       ...,\n       [1.51203990e+02, 2.34826122e-01],\n       [2.39833129e+02, 2.35251186e-01],\n       [1.49249027e+02, 3.04576535e-02]])"
     },
     "execution_count": 101,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_test_unscaled = scaler.inverse_transform(y_test)\n",
    "y_test_unscaled"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "outputs": [
    {
     "data": {
      "text/plain": "array([[1.8157640e+02, 6.4776659e-01],\n       [1.9712506e+02, 6.0968357e-01],\n       [1.5427470e+02, 3.1926131e-01],\n       ...,\n       [1.5607784e+02, 2.7125862e-01],\n       [2.2930032e+02, 1.9054066e-01],\n       [1.8682938e+02, 1.2644555e-01]], dtype=float32)"
     },
     "execution_count": 102,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred_unscaled = scaler.inverse_transform(y_pred)\n",
    "y_pred_unscaled"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Snow Density Performance: \n",
      "MAE:  23.718610978952434\n",
      "MSE:  1191.754187683344\n",
      "RMSE:  34.521792938422884\n",
      "R2:  0.5597049718555008\n",
      "VOD Performance: \n",
      "MAE:  0.1320375867482831\n",
      "MSE:  0.05647704427997352\n",
      "RMSE:  0.23764899385432608\n",
      "R2:  0.30884914539383856\n"
     ]
    }
   ],
   "source": [
    "#Snow density:\n",
    "mae = skm.mean_absolute_error(y_test_unscaled[:, 0], y_pred_unscaled[:, 0])\n",
    "mse = skm.mean_squared_error(y_test_unscaled[:, 0], y_pred_unscaled[:, 0])\n",
    "rmse = skm.mean_squared_error(y_test_unscaled[:, 0], y_pred_unscaled[:, 0], squared=False)\n",
    "r2 = skm.r2_score(y_test_unscaled[:, 0], y_pred_unscaled[:, 0])\n",
    "print(\"Snow Density Performance: \")\n",
    "print(\"MAE: \", mae)\n",
    "print(\"MSE: \", mse)\n",
    "print(\"RMSE: \", rmse)\n",
    "print(\"R2: \", r2)\n",
    "\n",
    "#VOD:\n",
    "mae2 = skm.mean_absolute_error(y_test_unscaled[:, 1], y_pred_unscaled[:, 1])\n",
    "mse2 = skm.mean_squared_error(y_test_unscaled[:, 1], y_pred_unscaled[:, 1])\n",
    "rmse2 = skm.mean_squared_error(y_test_unscaled[:, 1], y_pred_unscaled[:, 1], squared=False)\n",
    "r22 = skm.r2_score(y_test_unscaled[:, 1], y_pred_unscaled[:, 1])\n",
    "print(\"VOD Performance: \")\n",
    "print(\"MAE: \", mae2)\n",
    "print(\"MSE: \", mse2)\n",
    "print(\"RMSE: \", rmse2)\n",
    "print(\"R2: \", r22)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Training 80 Sparse"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(220426, 80, 7)\n",
      "(220426, 2)\n",
      "(53072, 80, 7)\n",
      "(53072, 2)\n"
     ]
    }
   ],
   "source": [
    "print(X_train.shape)  #(4352938, 60, 7), (3508883, 60, 7), (407693, 60, 7), 80: (220426, 80, 7)\n",
    "print(y_train.shape)\n",
    "print(X_val.shape)\n",
    "print(y_val.shape)\n",
    "\n",
    "batch_size = 500\n",
    "\n",
    "callbacks_list = [\n",
    "    tf.keras.callbacks.EarlyStopping(\n",
    "        monitor=\"val_loss\",\n",
    "        patience=3\n",
    "    ),\n",
    "    tf.keras.callbacks.ModelCheckpoint(\n",
    "        filepath=\"ModelCheckpoints80/model.{epoch:02d}-{val_loss:.2f}.h5\",\n",
    "        monitor=\"val_loss\",\n",
    "        mode=\"min\",\n",
    "        save_best_only=True,\n",
    "    )\n",
    "]"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 127,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " lstm_4 (LSTM)               (None, 80, 60)            16320     \n",
      "                                                                 \n",
      " dropout_4 (Dropout)         (None, 80, 60)            0         \n",
      "                                                                 \n",
      " lstm_5 (LSTM)               (None, 60)                29040     \n",
      "                                                                 \n",
      " dropout_5 (Dropout)         (None, 60)                0         \n",
      "                                                                 \n",
      " dense_2 (Dense)             (None, 2)                 122       \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 45,482\n",
      "Trainable params: 45,482\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model2 = model_lstm()\n",
    "model2.summary()"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Time:  18:17:37\n",
      "Epoch 1/20\n",
      "441/441 [==============================] - 386s 870ms/step - loss: 0.0217 - val_loss: 0.0199\n",
      "Epoch 2/20\n",
      "441/441 [==============================] - 380s 863ms/step - loss: 0.0173 - val_loss: 0.0149\n",
      "Epoch 3/20\n",
      "441/441 [==============================] - 381s 864ms/step - loss: 0.0158 - val_loss: 0.0143\n",
      "Epoch 4/20\n",
      "441/441 [==============================] - 363s 824ms/step - loss: 0.0147 - val_loss: 0.0137\n",
      "Epoch 5/20\n",
      "441/441 [==============================] - 376s 854ms/step - loss: 0.0140 - val_loss: 0.0126\n",
      "Epoch 6/20\n",
      "441/441 [==============================] - 376s 852ms/step - loss: 0.0138 - val_loss: 0.0121\n",
      "Epoch 7/20\n",
      "441/441 [==============================] - 385s 873ms/step - loss: 0.0132 - val_loss: 0.0140\n",
      "Epoch 8/20\n",
      "441/441 [==============================] - 390s 883ms/step - loss: 0.0128 - val_loss: 0.0136\n",
      "Epoch 9/20\n",
      "441/441 [==============================] - 379s 859ms/step - loss: 0.0127 - val_loss: 0.0115\n",
      "Epoch 10/20\n",
      "441/441 [==============================] - 383s 868ms/step - loss: 0.0121 - val_loss: 0.0128\n",
      "Epoch 11/20\n",
      "441/441 [==============================] - 387s 878ms/step - loss: 0.0116 - val_loss: 0.0106\n",
      "Epoch 12/20\n",
      "441/441 [==============================] - 387s 877ms/step - loss: 0.0114 - val_loss: 0.0103\n",
      "Epoch 13/20\n",
      "441/441 [==============================] - 388s 880ms/step - loss: 0.0114 - val_loss: 0.0102\n",
      "Epoch 14/20\n",
      "441/441 [==============================] - 379s 859ms/step - loss: 0.0110 - val_loss: 0.0099\n",
      "Epoch 15/20\n",
      "441/441 [==============================] - 383s 868ms/step - loss: 0.0109 - val_loss: 0.0101\n",
      "Epoch 16/20\n",
      "441/441 [==============================] - 382s 867ms/step - loss: 0.0108 - val_loss: 0.0099\n",
      "Epoch 17/20\n",
      "441/441 [==============================] - 377s 856ms/step - loss: 0.0106 - val_loss: 0.0098\n",
      "Epoch 18/20\n",
      "441/441 [==============================] - 382s 866ms/step - loss: 0.0105 - val_loss: 0.0095\n",
      "Epoch 19/20\n",
      "441/441 [==============================] - 387s 877ms/step - loss: 0.0103 - val_loss: 0.0098\n",
      "Epoch 20/20\n",
      "441/441 [==============================] - 405s 919ms/step - loss: 0.0103 - val_loss: 0.0094\n",
      "Time:  20:25:13\n"
     ]
    }
   ],
   "source": [
    "print(\"Time: \", time.strftime(\"%H:%M:%S\", time.localtime()))\n",
    "\n",
    "history_retrieval2 = model2.fit(X_train, y_train, epochs=20, validation_data=(X_val, y_val),\n",
    "                              batch_size=batch_size, callbacks=callbacks_list, verbose=1)\n",
    "\n",
    "print(\"Time: \", time.strftime(\"%H:%M:%S\", time.localtime()))"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 129,
   "outputs": [
    {
     "data": {
      "text/plain": "[<matplotlib.lines.Line2D at 0x23f58e8dfa0>]"
     },
     "execution_count": 129,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": "<Figure size 640x480 with 1 Axes>",
      "image/png": "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\n"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss_per_epoch = model2.history.history['loss']\n",
    "plt.plot(range(len(loss_per_epoch)), loss_per_epoch)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "outputs": [],
   "source": [
    "model2.save('ModelCheckpoints80/model80sparse_final.h5')"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "441/441 [==============================] - 83s 188ms/step - loss: 0.0098\n"
     ]
    },
    {
     "data": {
      "text/plain": "0.00978927593678236"
     },
     "execution_count": 131,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model2.evaluate(X_train, y_train, batch_size=batch_size, callbacks=callbacks_list, verbose=1)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "107/107 [==============================] - 21s 194ms/step - loss: 0.0094\n"
     ]
    },
    {
     "data": {
      "text/plain": "0.009396200999617577"
     },
     "execution_count": 132,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model2.evaluate(X_val, y_val, batch_size=batch_size, callbacks=callbacks_list, verbose=1)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_0.npy'> (2850, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_1.npy'> (5722, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_10.npy'> (8552, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_11.npy'> (11412, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_12.npy'> (14182, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_13.npy'> (17087, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_14.npy'> (19873, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_15.npy'> (22815, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_16.npy'> (25684, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_17.npy'> (28538, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_18.npy'> (31424, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_19.npy'> (34401, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_2.npy'> (37272, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_20.npy'> (40068, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_21.npy'> (42993, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_22.npy'> (45858, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_23.npy'> (48707, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_24.npy'> (51550, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_25.npy'> (54451, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_26.npy'> (57274, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_27.npy'> (60170, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_28.npy'> (63006, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_29.npy'> (65883, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_3.npy'> (68762, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_30.npy'> (71613, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_31.npy'> (74432, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_32.npy'> (77344, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_33.npy'> (80150, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_34.npy'> (83012, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_35.npy'> (85842, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_36.npy'> (88645, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_37.npy'> (91474, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_38.npy'> (94417, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_39.npy'> (97269, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_4.npy'> (100093, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_40.npy'> (103018, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_41.npy'> (105874, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_42.npy'> (108848, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_43.npy'> (111654, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_44.npy'> (114556, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_45.npy'> (117351, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_46.npy'> (120162, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_47.npy'> (123002, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_48.npy'> (125922, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_5.npy'> (128757, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_6.npy'> (131641, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_7.npy'> (134524, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_8.npy'> (137324, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_9.npy'> (140223, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_80Sparse_Test\\\\XY_arrays_80Sparse_LSTM_Test_last.npy'> (142225, 80, 7)\n"
     ]
    }
   ],
   "source": [
    "X_test = np.empty([0, n_input, n_features])\n",
    "y_test = np.empty([0,2])\n",
    "path = 'LSTM_Data_80Sparse_Test'\n",
    "X_test, y_test = assemble_test(X_test, y_test, path)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(142225, 80, 7)\n",
      "(142225, 2)\n"
     ]
    }
   ],
   "source": [
    "print(X_test.shape)\n",
    "print(y_test.shape)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 137,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "285/285 [==============================] - 56s 193ms/step\n"
     ]
    }
   ],
   "source": [
    "y_pred = model2.predict(X_test, batch_size=batch_size, callbacks=callbacks_list, verbose=1)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "outputs": [
    {
     "data": {
      "text/plain": "array([[1.8332982e+02, 1.6236126e-01],\n       [2.4451628e+02, 2.1423501e-01],\n       [2.0108072e+02, 1.6824916e-01],\n       ...,\n       [1.5654552e+02, 4.5745993e-01],\n       [1.9972498e+02, 5.1051116e-01],\n       [2.2410820e+02, 4.8677689e-01]], dtype=float32)"
     },
     "execution_count": 139,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_test_unscaled = scaler.inverse_transform(y_test)\n",
    "y_test_unscaled\n",
    "\n",
    "y_pred_unscaled = scaler.inverse_transform(y_pred)\n",
    "y_pred_unscaled"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Metrics for 80 sparse, scaled:\n",
      "Snow Density Performance: \n",
      "MAE:  0.06368325221758922\n",
      "MSE:  0.009844589015032493\n",
      "RMSE:  0.09921990231315739\n",
      "R2:  0.6366401227278846\n",
      "VOD Performance: \n",
      "MAE:  0.0704987100757687\n",
      "MSE:  0.016163048150139425\n",
      "RMSE:  0.12713397716637131\n",
      "R2:  0.31666451781676475\n"
     ]
    },
    {
     "data": {
      "text/plain": "(0.06368325221758922,\n 0.009844589015032493,\n 0.09921990231315739,\n 0.6366401227278846,\n 0.0704987100757687,\n 0.016163048150139425,\n 0.12713397716637131,\n 0.31666451781676475)"
     },
     "execution_count": 138,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(\"Metrics for 80 sparse, scaled:\")\n",
    "get_metrics(y_test, y_pred)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Metrics for 80 sparse, unscaled:\n",
      "Snow Density Performance: \n",
      "MAE:  22.289138023393658\n",
      "MSE:  1205.9621534868168\n",
      "RMSE:  34.726965797299606\n",
      "R2:  0.636640122985397\n",
      "VOD Performance: \n",
      "MAE:  0.10573912753212986\n",
      "MSE:  0.036360711321544385\n",
      "RMSE:  0.19068484816981235\n",
      "R2:  0.3166645178093449\n"
     ]
    },
    {
     "data": {
      "text/plain": "(22.289138023393658,\n 1205.9621534868168,\n 34.726965797299606,\n 0.636640122985397,\n 0.10573912753212986,\n 0.036360711321544385,\n 0.19068484816981235,\n 0.3166645178093449)"
     },
     "execution_count": 140,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(\"Metrics for 80 sparse, unscaled:\")\n",
    "get_metrics(y_test_unscaled, y_pred_unscaled)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Data Prep Sparse 30 Train"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 153,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Length of df of unique latlon paris that meet cutoff:  215553\n",
      "Counter:  30000 , Shape X:  (48892, 30, 7) 2022-12-20 21:12:33\n",
      "Wipeout:  2\n",
      "Counter:  40000 , Shape X:  (49102, 30, 7) 2022-12-20 21:20:47\n",
      "Wipeout:  3\n",
      "Counter:  50000 , Shape X:  (48974, 30, 7) 2022-12-20 21:29:03\n",
      "Wipeout:  4\n",
      "Counter:  60000 , Shape X:  (49356, 30, 7) 2022-12-20 21:37:01\n",
      "Wipeout:  5\n",
      "Counter:  70000 , Shape X:  (49228, 30, 7) 2022-12-20 21:45:10\n",
      "Wipeout:  6\n",
      "Counter:  80000 , Shape X:  (48560, 30, 7) 2022-12-20 21:53:17\n",
      "Wipeout:  7\n",
      "Counter:  90000 , Shape X:  (48654, 30, 7) 2022-12-20 22:01:30\n",
      "Wipeout:  8\n",
      "Counter:  100000 , Shape X:  (48670, 30, 7) 2022-12-20 22:09:40\n",
      "Wipeout:  9\n",
      "Counter:  110000 , Shape X:  (49140, 30, 7) 2022-12-20 22:17:49\n",
      "Wipeout:  10\n",
      "Counter:  120000 , Shape X:  (49625, 30, 7) 2022-12-20 22:26:08\n",
      "Wipeout:  11\n",
      "Counter:  130000 , Shape X:  (48327, 30, 7) 2022-12-20 22:34:34\n",
      "Wipeout:  12\n",
      "Counter:  140000 , Shape X:  (48650, 30, 7) 2022-12-20 22:43:02\n",
      "Wipeout:  13\n",
      "Counter:  150000 , Shape X:  (48753, 30, 7) 2022-12-20 22:51:51\n",
      "Wipeout:  14\n",
      "Counter:  160000 , Shape X:  (49271, 30, 7) 2022-12-20 23:00:13\n",
      "Wipeout:  15\n",
      "Counter:  170000 , Shape X:  (48819, 30, 7) 2022-12-20 23:08:42\n",
      "Wipeout:  16\n",
      "Counter:  180000 , Shape X:  (48865, 30, 7) 2022-12-20 23:16:46\n",
      "Wipeout:  17\n",
      "Counter:  190000 , Shape X:  (48725, 30, 7) 2022-12-20 23:25:00\n",
      "Wipeout:  18\n",
      "Counter:  200000 , Shape X:  (48835, 30, 7) 2022-12-20 23:33:04\n",
      "Wipeout:  19\n",
      "Counter:  210000 , Shape X:  (48772, 30, 7) 2022-12-20 23:41:06\n",
      "Wipeout:  20\n",
      "Save last: \n",
      "Done! And it only took from  2022-12-20 21:03:00  to  2022-12-20 23:45:16\n"
     ]
    }
   ],
   "source": [
    "n_input = 30\n",
    "n_features = 7\n",
    "latlon = latlon_setup(train, n_input)\n",
    "X = np.empty([0, n_input, n_features])\n",
    "y = np.empty([0,2])\n",
    "path = 'LSTM_Data_30Sparse_Train/XY_arrays_30Sparse_LSTM_Train_'\n",
    "counter = 20000\n",
    "savept = 2\n",
    "\n",
    "data_prep(train, latlon, path, n_input, n_features, X, y, counter, savept)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 155,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LSTM_Data_30Sparse_Train\\XY_arrays_30Sparse_LSTM_Train0.npy\n",
      "Train:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train0.npy'>\n",
      "LSTM_Data_30Sparse_Train\\XY_arrays_30Sparse_LSTM_Train1.npy\n",
      "Train:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train1.npy'>\n",
      "LSTM_Data_30Sparse_Train\\XY_arrays_30Sparse_LSTM_Train_10.npy\n",
      "Train:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train_10.npy'>\n",
      "LSTM_Data_30Sparse_Train\\XY_arrays_30Sparse_LSTM_Train_11.npy\n",
      "Train:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train_11.npy'>\n",
      "LSTM_Data_30Sparse_Train\\XY_arrays_30Sparse_LSTM_Train_12.npy\n",
      "Train:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train_12.npy'>\n",
      "LSTM_Data_30Sparse_Train\\XY_arrays_30Sparse_LSTM_Train_13.npy\n",
      "Train:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train_13.npy'>\n",
      "LSTM_Data_30Sparse_Train\\XY_arrays_30Sparse_LSTM_Train_14.npy\n",
      "Train:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train_14.npy'>\n",
      "LSTM_Data_30Sparse_Train\\XY_arrays_30Sparse_LSTM_Train_15.npy\n",
      "Train:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train_15.npy'>\n",
      "LSTM_Data_30Sparse_Train\\XY_arrays_30Sparse_LSTM_Train_16.npy\n",
      "Train:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train_16.npy'>\n",
      "LSTM_Data_30Sparse_Train\\XY_arrays_30Sparse_LSTM_Train_17.npy\n",
      "Train:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train_17.npy'>\n",
      "LSTM_Data_30Sparse_Train\\XY_arrays_30Sparse_LSTM_Train_18.npy\n",
      "Train:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train_18.npy'>\n",
      "LSTM_Data_30Sparse_Train\\XY_arrays_30Sparse_LSTM_Train_19.npy\n",
      "Train:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train_19.npy'>\n",
      "LSTM_Data_30Sparse_Train\\XY_arrays_30Sparse_LSTM_Train_2.npy\n",
      "Train:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train_2.npy'>\n",
      "LSTM_Data_30Sparse_Train\\XY_arrays_30Sparse_LSTM_Train_20.npy\n",
      "Train:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train_20.npy'>\n",
      "LSTM_Data_30Sparse_Train\\XY_arrays_30Sparse_LSTM_Train_3.npy\n",
      "Train:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train_3.npy'>\n",
      "LSTM_Data_30Sparse_Train\\XY_arrays_30Sparse_LSTM_Train_4.npy\n",
      "Train:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train_4.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train_5.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train_6.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train_7.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train_8.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train_9.npy'>\n",
      "Val:  <_io.BufferedReader name='LSTM_Data_30Sparse_Train\\\\XY_arrays_30Sparse_LSTM_Train__last.npy'>\n"
     ]
    }
   ],
   "source": [
    "X_train = np.empty([0, n_input, n_features])\n",
    "y_train = np.empty([0,2])\n",
    "X_val = np.empty([0, n_input, n_features])\n",
    "y_val = np.empty([0,2])\n",
    "path = 'LSTM_Data_30Sparse_Train'\n",
    "switch = int(len(os.listdir(path))*.2)\n",
    "X_train, y_train, X_val, y_val = assemble_train(X_train, y_train, X_val, y_val, path, switch)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(782744, 30, 7)\n",
      "(782744, 2)\n",
      "(271881, 30, 7)\n",
      "(271881, 2)\n",
      "0.2577987436292521\n"
     ]
    }
   ],
   "source": [
    "print(X_train.shape) #(4352938, 60, 7), 30:(782744, 30, 7)\n",
    "print(y_train.shape)\n",
    "print(X_val.shape)\n",
    "print(y_val.shape)\n",
    "print(X_val.shape[0] / (X_val.shape[0] + X_train.shape[0])) #60: 0.192, 80: 0.194\n",
    "with open((path + '/XY_30Sparse_Full_XYTrainVal.npy'), 'wb') as f:\n",
    "      np.save(f, X_train)\n",
    "      np.save(f, y_train)\n",
    "      np.save(f, X_val)\n",
    "      np.save(f, y_val)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 157,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_3\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " lstm_6 (LSTM)               (None, 30, 60)            16320     \n",
      "                                                                 \n",
      " dropout_6 (Dropout)         (None, 30, 60)            0         \n",
      "                                                                 \n",
      " lstm_7 (LSTM)               (None, 60)                29040     \n",
      "                                                                 \n",
      " dropout_7 (Dropout)         (None, 60)                0         \n",
      "                                                                 \n",
      " dense_3 (Dense)             (None, 2)                 122       \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 45,482\n",
      "Trainable params: 45,482\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = model_lstm()\n",
    "model.summary()"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 159,
   "outputs": [],
   "source": [
    "batch_size = 500\n",
    "\n",
    "callbacks_list = [\n",
    "    tf.keras.callbacks.EarlyStopping(\n",
    "        monitor=\"val_loss\",\n",
    "        patience=3\n",
    "    ),\n",
    "    tf.keras.callbacks.ModelCheckpoint(\n",
    "        filepath=\"ModelCheckpoints30/model.{epoch:02d}-{val_loss:.2f}.h5\",\n",
    "        monitor=\"val_loss\",\n",
    "        mode=\"min\",\n",
    "        save_best_only=True,\n",
    "    )\n",
    "]"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 160,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Time:  23:57:15\n",
      "Epoch 1/20\n",
      "1566/1566 [==============================] - 372s 236ms/step - loss: 0.0243 - val_loss: 0.0197\n",
      "Epoch 2/20\n",
      "1566/1566 [==============================] - 345s 220ms/step - loss: 0.0195 - val_loss: 0.0180\n",
      "Epoch 3/20\n",
      "1566/1566 [==============================] - 273s 175ms/step - loss: 0.0183 - val_loss: 0.0182\n",
      "Epoch 4/20\n",
      "1566/1566 [==============================] - 284s 181ms/step - loss: 0.0175 - val_loss: 0.0168\n",
      "Epoch 5/20\n",
      "1566/1566 [==============================] - 301s 192ms/step - loss: 0.0170 - val_loss: 0.0171\n",
      "Epoch 6/20\n",
      "1566/1566 [==============================] - 359s 229ms/step - loss: 0.0165 - val_loss: 0.0159\n",
      "Epoch 7/20\n",
      "1566/1566 [==============================] - 428s 273ms/step - loss: 0.0162 - val_loss: 0.0162\n",
      "Epoch 8/20\n",
      "1566/1566 [==============================] - 440s 281ms/step - loss: 0.0158 - val_loss: 0.0157\n",
      "Epoch 9/20\n",
      "1566/1566 [==============================] - 444s 284ms/step - loss: 0.0156 - val_loss: 0.0144\n",
      "Epoch 10/20\n",
      "1566/1566 [==============================] - 459s 293ms/step - loss: 0.0153 - val_loss: 0.0143\n",
      "Epoch 11/20\n",
      "1566/1566 [==============================] - 465s 297ms/step - loss: 0.0150 - val_loss: 0.0156\n",
      "Epoch 12/20\n",
      "1566/1566 [==============================] - 474s 303ms/step - loss: 0.0148 - val_loss: 0.0143\n",
      "Epoch 13/20\n",
      "1566/1566 [==============================] - 478s 305ms/step - loss: 0.0147 - val_loss: 0.0140\n",
      "Epoch 14/20\n",
      "1566/1566 [==============================] - 490s 313ms/step - loss: 0.0145 - val_loss: 0.0144\n",
      "Epoch 15/20\n",
      "1566/1566 [==============================] - 492s 314ms/step - loss: 0.0143 - val_loss: 0.0140\n",
      "Epoch 16/20\n",
      "1566/1566 [==============================] - 499s 318ms/step - loss: 0.0142 - val_loss: 0.0134\n",
      "Epoch 17/20\n",
      "1566/1566 [==============================] - 495s 316ms/step - loss: 0.0141 - val_loss: 0.0135\n",
      "Epoch 18/20\n",
      "1566/1566 [==============================] - 496s 317ms/step - loss: 0.0140 - val_loss: 0.0137\n",
      "Epoch 19/20\n",
      "1566/1566 [==============================] - 496s 317ms/step - loss: 0.0138 - val_loss: 0.0137\n",
      "Time:  02:12:05\n"
     ]
    }
   ],
   "source": [
    "print(\"Time: \", time.strftime(\"%H:%M:%S\", time.localtime()))\n",
    "\n",
    "history_retrieval = model.fit(X_train, y_train, epochs=20, validation_data=(X_val, y_val),\n",
    "                              batch_size=batch_size, callbacks=callbacks_list, verbose=1)\n",
    "\n",
    "print(\"Time: \", time.strftime(\"%H:%M:%S\", time.localtime()))\n",
    "\n"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 161,
   "outputs": [],
   "source": [
    "model.save('ModelCheckpoints30/model30sparse_final.h5')"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "outputs": [
    {
     "data": {
      "text/plain": "[<matplotlib.lines.Line2D at 0x23f59953af0>]"
     },
     "execution_count": 162,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": "<Figure size 640x480 with 1 Axes>",
      "image/png": "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\n"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss_per_epoch = model.history.history['loss']\n",
    "plt.plot(range(len(loss_per_epoch)), loss_per_epoch)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 163,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1566/1566 [==============================] - 134s 85ms/step - loss: 0.0135\n"
     ]
    },
    {
     "data": {
      "text/plain": "0.013468669727444649"
     },
     "execution_count": 163,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(X_train, y_train, batch_size=batch_size, callbacks=callbacks_list, verbose=1)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 164,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "544/544 [==============================] - 49s 89ms/step - loss: 0.0301\n"
     ]
    },
    {
     "data": {
      "text/plain": "0.030079761520028114"
     },
     "execution_count": 164,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model2.evaluate(X_val, y_val, batch_size=batch_size, callbacks=callbacks_list, verbose=1)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 165,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_0.npy'> (48892, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_1.npy'> (97994, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_10.npy'> (146321, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_11.npy'> (194971, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_12.npy'> (243724, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_13.npy'> (292995, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_14.npy'> (341814, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_15.npy'> (390679, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_16.npy'> (439404, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_17.npy'> (488239, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_18.npy'> (537011, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_19.npy'> (585936, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_2.npy'> (634910, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_20.npy'> (683934, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_3.npy'> (733290, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_4.npy'> (782518, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_5.npy'> (831078, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_6.npy'> (879732, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_7.npy'> (928402, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_8.npy'> (977542, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test_9.npy'> (1027167, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Data_30Sparse_Test\\\\XY_arrays_30Sparse_LSTM_Test__last.npy'> (1054300, 30, 7)\n",
      "(1054300, 30, 7)\n",
      "(1054300, 2)\n"
     ]
    }
   ],
   "source": [
    "X_test = np.empty([0, n_input, n_features])\n",
    "y_test = np.empty([0, 2])\n",
    "path = 'LSTM_Data_30Sparse_Test'\n",
    "X_test, y_test = assemble_test(X_test, y_test, path)\n",
    "\n",
    "print(X_test.shape)\n",
    "print(y_test.shape)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 166,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2109/2109 [==============================] - 219s 103ms/step\n"
     ]
    }
   ],
   "source": [
    "y_pred = model.predict(X_test, batch_size=batch_size, callbacks=callbacks_list, verbose=1)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "outputs": [
    {
     "data": {
      "text/plain": "array([[2.9794083e+02, 2.6936841e-01],\n       [2.5505013e+02, 1.9362469e-01],\n       [2.3156006e+02, 2.0710987e-01],\n       ...,\n       [1.6291206e+02, 3.9036390e-01],\n       [1.7536044e+02, 3.4844720e-01],\n       [1.7036363e+02, 3.3901000e-01]], dtype=float32)"
     },
     "execution_count": 167,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_test_unscaled = scaler.inverse_transform(y_test)\n",
    "y_test_unscaled\n",
    "\n",
    "y_pred_unscaled = scaler.inverse_transform(y_pred)\n",
    "y_pred_unscaled"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 168,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Metrics for 80 sparse, scaled:\n",
      "Snow Density Performance: \n",
      "MAE:  0.03625762413142435\n",
      "MSE:  0.0035934464750510043\n",
      "RMSE:  0.05994536241487747\n",
      "R2:  0.7628973558285523\n",
      "VOD Performance: \n",
      "MAE:  0.08452074440335217\n",
      "MSE:  0.023705100653086925\n",
      "RMSE:  0.15396460844326182\n",
      "R2:  0.6513472458784648\n",
      "Metrics for 80 sparse, unscaled:\n",
      "Snow Density Performance: \n",
      "MAE:  12.690168333240559\n",
      "MSE:  440.19718218177934\n",
      "RMSE:  20.980876582778407\n",
      "R2:  0.7628973617599096\n",
      "VOD Performance: \n",
      "MAE:  0.1267704013672531\n",
      "MSE:  0.05332746110486599\n",
      "RMSE:  0.23092739357829767\n",
      "R2:  0.6513472459282392\n"
     ]
    },
    {
     "data": {
      "text/plain": "(12.690168333240559,\n 440.19718218177934,\n 20.980876582778407,\n 0.7628973617599096,\n 0.1267704013672531,\n 0.05332746110486599,\n 0.23092739357829767,\n 0.6513472459282392)"
     },
     "execution_count": 168,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(\"Metrics for 80 sparse, scaled:\")\n",
    "get_metrics(y_test, y_pred)\n",
    "print(\"Metrics for 80 sparse, unscaled:\")\n",
    "get_metrics(y_test_unscaled, y_pred_unscaled)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 241,
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\calvino\\Documents\\Courses\\GradSchoolStuff\\CSCI 8523 - AI for Earth, Kumar\\Group Project\\venv\\lib\\site-packages\\mpl_scatter_density\\generic_density_artist.py:77: RuntimeWarning: All-NaN slice encountered\n",
      "  vmin = self._density_vmin(array)\n",
      "C:\\Users\\calvino\\Documents\\Courses\\GradSchoolStuff\\CSCI 8523 - AI for Earth, Kumar\\Group Project\\venv\\lib\\site-packages\\mpl_scatter_density\\generic_density_artist.py:82: RuntimeWarning: All-NaN slice encountered\n",
      "  vmax = self._density_vmax(array)\n"
     ]
    },
    {
     "data": {
      "text/plain": "<Figure size 640x480 with 2 Axes>",
      "image/png": "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\n"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#y_test = y_test.to_numpy()\n",
    "\n",
    "fig = plt.figure()\n",
    "\n",
    "white_viridis = LinearSegmentedColormap.from_list('white_viridis', [\n",
    "    (0, '#ffffff'),\n",
    "    (1e-20, '#440053'),\n",
    "    (0.2, '#404388'),\n",
    "    (0.4, '#2a788e'),\n",
    "    (0.6, '#21a784'),\n",
    "    (0.8, '#78d151'),\n",
    "    (1, '#fde624'),\n",
    "], N=256)\n",
    "\n",
    "ax = fig.add_subplot(1, 1, 1, projection='scatter_density')\n",
    "density = ax.scatter_density(y_test, y_pred, cmap=white_viridis, vmin=0, vmax=10000)\n",
    "fig.colorbar(density, label='Number of points per pixel')\n",
    "plt.show()"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Map best performer for SD:"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 170,
   "outputs": [],
   "source": [
    "#First we need to pick a day to map:\n",
    "\n",
    "test = pd.read_parquet('test_scaled_LatLonDateUnscaled.parquet')"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 190,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Length of df of unique latlon paris that meet cutoff:  153229\n",
      "              lat        lon\n",
      "139568  65.905140   95.74170\n",
      "5627    33.165960   78.56328\n",
      "96845   58.551846  131.31224\n",
      "115202  61.636730 -150.17116\n",
      "73378   55.222220  -76.41598\n",
      "Length of testdate:  58314\n",
      "           density  snowtemp  soiltemp     omega       Tbh       Tbv  \\\n",
      "30586952  0.170206  0.708772  0.845780  0.681356  0.705369  0.713866   \n",
      "30586953  0.167071  0.730467  0.868891  0.680220  0.688666  0.689544   \n",
      "30586954  0.159643  0.697934  0.845582  0.680000  0.690524  0.704247   \n",
      "30586955  0.170206  0.708772  0.857352  0.681035  0.701983  0.712883   \n",
      "30586956  0.174646  0.704454  0.869552  0.700000  0.698099  0.709491   \n",
      "\n",
      "                cf         h       voc        lat        lon       date  \n",
      "30586952  0.323790  0.090114  0.064464  65.565310 -179.95332 2020-11-01  \n",
      "30586953  0.322650  0.085397  0.004672  65.229836 -179.95332 2020-11-01  \n",
      "30586954  0.314897  0.095943  0.068366  65.734665 -179.85995 2020-11-01  \n",
      "30586955  0.336332  0.088294  0.065556  65.565310 -179.85995 2020-11-01  \n",
      "30586956  0.353848  0.088024  0.064836  65.565310 -179.76660 2020-11-01  \n",
      "Length of testdate after merge:  42596\n",
      "    density  snowtemp  soiltemp     omega       Tbh       Tbv        cf  \\\n",
      "0  0.152719  0.710146  0.995139  0.700000  0.721610  0.726299  0.330263   \n",
      "1  0.145295  0.690952  0.985973  0.700826  0.719028  0.717893  0.288278   \n",
      "2  0.145243  0.653960  0.993925  0.700000  0.710364  0.713868  0.290911   \n",
      "3  0.145570  0.698852  0.985099  0.702479  0.721501  0.719431  0.275937   \n",
      "4  0.144965  0.689153  0.976180  0.700826  0.717999  0.721260  0.282593   \n",
      "\n",
      "          h       voc        lat        lon       date  \n",
      "0  0.091574  0.078173  68.047226 -163.89523 2020-11-01  \n",
      "1  0.106283  0.074326  67.676765 -162.40146 2020-11-01  \n",
      "2  0.104408  0.072966  67.493700 -162.40146 2020-11-01  \n",
      "3  0.112064  0.082468  67.861270 -162.30809 2020-11-01  \n",
      "4  0.107667  0.076295  67.676765 -162.30809 2020-11-01  \n",
      "Final len testdate:  42596\n",
      "Final len latlon1:  42596\n"
     ]
    }
   ],
   "source": [
    "test1 = test[test['date'] <= '2020-11-1'] #All records up to our map day\n",
    "\n",
    "latlon1 = latlon_setup(test1, 30)\n",
    "print(latlon1.head()) #Maximum possible pixels.\n",
    "\n",
    "testdate = test[test['date'] == '2020-11-1'] #our y_test, essentially.\n",
    "print(\"Length of testdate: \", len(testdate))\n",
    "print(testdate.head()) #The list of pixels present on map day\n",
    "\n",
    "#Only keep pixels on testdate that are also present in latlon1:\n",
    "testdate = pd.merge(testdate, latlon1, how = 'inner', on = ['lat', 'lon'])\n",
    "latlon1 = pd.merge(latlon1, testdate, how = 'inner', on = ['lat', 'lon'])\n",
    "print(\"Length of testdate after merge: \", len(testdate))\n",
    "print(testdate.head())\n",
    "print(\"Final len testdate: \", len(testdate)) #80: 19681, 30: 42596\n",
    "print(\"Final len latlon1: \", len(latlon1))"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 218,
   "outputs": [],
   "source": [
    "def gen_np_last(df, n_input):\n",
    "  \"\"\"Produces last possible window of n_input historical records of data from the provided dataframe.\"\"\"\n",
    "  Xs, ys = [], []\n",
    "  current_slice = df.iloc[len(df)-n_input:]\n",
    "  X = current_slice[['snowtemp', 'soiltemp', 'omega', 'Tbh', 'Tbv', 'cf', 'h']].values\n",
    "  y = current_slice[['density', 'voc']].values[-1]\n",
    "  Xs.append(X)\n",
    "  ys.append(y)\n",
    "  X = np.stack(Xs)\n",
    "  y = np.stack(ys)\n",
    "  return X, y"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 222,
   "outputs": [],
   "source": [
    "#Iterate though test and grab the LAST possible sample set of length n_length, such that the last X values and the corresponding y values are in testdate.\n",
    "def data_prep_last_only(bigdf, latlon, path, n_input, n_features, X, y, counter = 0, savept = 0):\n",
    "    start = time.strftime(\"%Y-%m-%d %H:%M:%S\", time.localtime())\n",
    "    print(\"Started at: \", start)\n",
    "    for i in range(len(latlon)-counter): #Iterate though our set of viable lat/lons, minus counter to restart if needed.\n",
    "      counter += 1\n",
    "      df = bigdf[(bigdf['lat'] == latlon['lat'].iloc[i]) & (bigdf['lon'] == latlon['lon'].iloc[i])] #Make a sub df of each lat/lon pair\n",
    "      df = df.sort_values(by = 'date') #Make sure we have our values correctly ordered\n",
    "      #print(\"Len of sub df: \", len(df))\n",
    "      X2, y2 = gen_np_last(df, n_input) #Make your arrays, in this case only last possible array such that they end on a specific date.\n",
    "      #print(\"Dimensions of X and y: \", X.shape, y.shape)\n",
    "      #print(\"Dimensions of X2 and y2: \", X2.shape, y2.shape)\n",
    "      X = np.vstack([X, X2]) #Append your arrays\n",
    "      y = np.concatenate([y, y2])\n",
    "      if counter%1000 == 0: #For every thousand lat/long pairs check point\n",
    "        print(\"Counter: \", counter, \", Shape X: \", X.shape, time.strftime(\"%Y-%m-%d %H:%M:%S\", time.localtime()))\n",
    "        with open((path + str(savept) + '.npy'), 'wb') as f: #Saves two arrays in same file, arrays can be opened separately.\n",
    "          np.save(f, X)\n",
    "          np.save(f, y)\n",
    "        print(\"Wipeout: \", savept)\n",
    "        savept += 1\n",
    "        X = np.empty([0, n_input, n_features])\n",
    "        y = np.empty([0,2])\n",
    "    print(\"Save last: \")\n",
    "    with open((path + '_last.npy'), 'wb') as f: #final save of X,y to grab any remaining lat/lon pairs since last checkpoint:\n",
    "          np.save(f, X)\n",
    "          np.save(f, y)\n",
    "    print(\"Done! And it only took from \", start, \" to \", time.strftime(\"%Y-%m-%d %H:%M:%S\", time.localtime()))\n"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 223,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Started at:  2022-12-21 16:39:34\n",
      "Counter:  1000 , Shape X:  (1000, 30, 7) 2022-12-21 16:40:01\n",
      "Wipeout:  0\n",
      "Counter:  2000 , Shape X:  (1000, 30, 7) 2022-12-21 16:40:25\n",
      "Wipeout:  1\n",
      "Counter:  3000 , Shape X:  (1000, 30, 7) 2022-12-21 16:40:50\n",
      "Wipeout:  2\n",
      "Counter:  4000 , Shape X:  (1000, 30, 7) 2022-12-21 16:41:15\n",
      "Wipeout:  3\n",
      "Counter:  5000 , Shape X:  (1000, 30, 7) 2022-12-21 16:41:39\n",
      "Wipeout:  4\n",
      "Counter:  6000 , Shape X:  (1000, 30, 7) 2022-12-21 16:42:06\n",
      "Wipeout:  5\n",
      "Counter:  7000 , Shape X:  (1000, 30, 7) 2022-12-21 16:42:38\n",
      "Wipeout:  6\n",
      "Counter:  8000 , Shape X:  (1000, 30, 7) 2022-12-21 16:43:09\n",
      "Wipeout:  7\n",
      "Counter:  9000 , Shape X:  (1000, 30, 7) 2022-12-21 16:43:34\n",
      "Wipeout:  8\n",
      "Counter:  10000 , Shape X:  (1000, 30, 7) 2022-12-21 16:43:59\n",
      "Wipeout:  9\n",
      "Counter:  11000 , Shape X:  (1000, 30, 7) 2022-12-21 16:44:24\n",
      "Wipeout:  10\n",
      "Counter:  12000 , Shape X:  (1000, 30, 7) 2022-12-21 16:44:49\n",
      "Wipeout:  11\n",
      "Counter:  13000 , Shape X:  (1000, 30, 7) 2022-12-21 16:45:14\n",
      "Wipeout:  12\n",
      "Counter:  14000 , Shape X:  (1000, 30, 7) 2022-12-21 16:45:40\n",
      "Wipeout:  13\n",
      "Counter:  15000 , Shape X:  (1000, 30, 7) 2022-12-21 16:46:07\n",
      "Wipeout:  14\n",
      "Counter:  16000 , Shape X:  (1000, 30, 7) 2022-12-21 16:46:32\n",
      "Wipeout:  15\n",
      "Counter:  17000 , Shape X:  (1000, 30, 7) 2022-12-21 16:46:57\n",
      "Wipeout:  16\n",
      "Counter:  18000 , Shape X:  (1000, 30, 7) 2022-12-21 16:47:22\n",
      "Wipeout:  17\n",
      "Counter:  19000 , Shape X:  (1000, 30, 7) 2022-12-21 16:47:49\n",
      "Wipeout:  18\n",
      "Counter:  20000 , Shape X:  (1000, 30, 7) 2022-12-21 16:48:17\n",
      "Wipeout:  19\n",
      "Counter:  21000 , Shape X:  (1000, 30, 7) 2022-12-21 16:48:43\n",
      "Wipeout:  20\n",
      "Counter:  22000 , Shape X:  (1000, 30, 7) 2022-12-21 16:49:12\n",
      "Wipeout:  21\n",
      "Counter:  23000 , Shape X:  (1000, 30, 7) 2022-12-21 16:49:38\n",
      "Wipeout:  22\n",
      "Counter:  24000 , Shape X:  (1000, 30, 7) 2022-12-21 16:50:08\n",
      "Wipeout:  23\n",
      "Counter:  25000 , Shape X:  (1000, 30, 7) 2022-12-21 16:50:36\n",
      "Wipeout:  24\n",
      "Counter:  26000 , Shape X:  (1000, 30, 7) 2022-12-21 16:51:03\n",
      "Wipeout:  25\n",
      "Counter:  27000 , Shape X:  (1000, 30, 7) 2022-12-21 16:51:30\n",
      "Wipeout:  26\n",
      "Counter:  28000 , Shape X:  (1000, 30, 7) 2022-12-21 16:51:57\n",
      "Wipeout:  27\n",
      "Counter:  29000 , Shape X:  (1000, 30, 7) 2022-12-21 16:52:24\n",
      "Wipeout:  28\n",
      "Counter:  30000 , Shape X:  (1000, 30, 7) 2022-12-21 16:52:51\n",
      "Wipeout:  29\n",
      "Counter:  31000 , Shape X:  (1000, 30, 7) 2022-12-21 16:53:17\n",
      "Wipeout:  30\n",
      "Counter:  32000 , Shape X:  (1000, 30, 7) 2022-12-21 16:53:44\n",
      "Wipeout:  31\n",
      "Counter:  33000 , Shape X:  (1000, 30, 7) 2022-12-21 16:54:10\n",
      "Wipeout:  32\n",
      "Counter:  34000 , Shape X:  (1000, 30, 7) 2022-12-21 16:54:36\n",
      "Wipeout:  33\n",
      "Counter:  35000 , Shape X:  (1000, 30, 7) 2022-12-21 16:55:03\n",
      "Wipeout:  34\n",
      "Counter:  36000 , Shape X:  (1000, 30, 7) 2022-12-21 16:55:31\n",
      "Wipeout:  35\n",
      "Counter:  37000 , Shape X:  (1000, 30, 7) 2022-12-21 16:55:57\n",
      "Wipeout:  36\n",
      "Counter:  38000 , Shape X:  (1000, 30, 7) 2022-12-21 16:56:24\n",
      "Wipeout:  37\n",
      "Counter:  39000 , Shape X:  (1000, 30, 7) 2022-12-21 16:56:49\n",
      "Wipeout:  38\n",
      "Counter:  40000 , Shape X:  (1000, 30, 7) 2022-12-21 16:57:14\n",
      "Wipeout:  39\n",
      "Counter:  41000 , Shape X:  (1000, 30, 7) 2022-12-21 16:57:38\n",
      "Wipeout:  40\n",
      "Counter:  42000 , Shape X:  (1000, 30, 7) 2022-12-21 16:58:03\n",
      "Wipeout:  41\n",
      "Save last: \n",
      "Done! And it only took from  2022-12-21 16:39:34  to  2022-12-21 16:58:17\n"
     ]
    }
   ],
   "source": [
    "n_input = 30\n",
    "n_features = 7\n",
    "X = np.empty([0, n_input, n_features])\n",
    "y = np.empty([0,2])\n",
    "#print(X.shape)\n",
    "#print(y.shape)\n",
    "path = 'LSTM_Map_30'\n",
    "counter = 0\n",
    "savept = 0\n",
    "data_prep_last_only(test1, latlon1, path, n_input, n_features, X, y, counter, savept)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 224,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_300.npy'> (1000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_301.npy'> (2000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3010.npy'> (3000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3011.npy'> (4000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3012.npy'> (5000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3013.npy'> (6000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3014.npy'> (7000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3015.npy'> (8000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3016.npy'> (9000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3017.npy'> (10000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3018.npy'> (11000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3019.npy'> (12000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_302.npy'> (13000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3020.npy'> (14000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3021.npy'> (15000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3022.npy'> (16000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3023.npy'> (17000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3024.npy'> (18000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3025.npy'> (19000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3026.npy'> (20000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3027.npy'> (21000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3028.npy'> (22000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3029.npy'> (23000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_303.npy'> (24000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3030.npy'> (25000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3031.npy'> (26000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3032.npy'> (27000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3033.npy'> (28000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3034.npy'> (29000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3035.npy'> (30000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3036.npy'> (31000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3037.npy'> (32000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3038.npy'> (33000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3039.npy'> (34000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_304.npy'> (35000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3040.npy'> (36000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_3041.npy'> (37000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_305.npy'> (38000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_306.npy'> (39000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_307.npy'> (40000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_308.npy'> (41000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_309.npy'> (42000, 30, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_30\\\\LSTM_Map_30_last.npy'> (42596, 30, 7)\n",
      "(42596, 30, 7)\n",
      "(42596, 2)\n",
      "[[0.07724035 0.        ]\n",
      " [0.14380287 0.2548425 ]\n",
      " [0.14597607 0.19710033]\n",
      " [0.14330185 0.        ]\n",
      " [0.14332537 0.        ]]\n"
     ]
    }
   ],
   "source": [
    "X_map = np.empty([0, n_input, n_features])\n",
    "y_map = np.empty([0, 2])\n",
    "path = 'LSTM_Map_30'\n",
    "X_map, y_map = assemble_test(X_map, y_map, path)\n",
    "\n",
    "print(X_map.shape)\n",
    "print(y_map.shape)\n",
    "print(y_map[0:5])"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 225,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "86/86 [==============================] - 11s 123ms/step\n"
     ]
    }
   ],
   "source": [
    "y_pred_map = model.predict(X_map, batch_size=batch_size, callbacks=callbacks_list, verbose=1)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 226,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[127.03410623   0.        ]\n",
      " [150.33098846   0.38223145]\n",
      " [151.09161033   0.2956255 ]\n",
      " [150.15563151   0.        ]\n",
      " [150.16386528   0.        ]]\n",
      "[[ 1.5168173e+02 -5.5694939e-03]\n",
      " [ 1.6618825e+02  4.9331778e-01]\n",
      " [ 1.5827963e+02  5.6985205e-01]\n",
      " [ 1.5202991e+02 -8.3786920e-03]\n",
      " [ 1.5320758e+02  1.5642662e-03]]\n",
      "Metrics for nov 1st 2020 map, unscaled:\n",
      "Snow Density Performance: \n",
      "MAE:  15.512991386137019\n",
      "MSE:  632.1912049561365\n",
      "RMSE:  25.143412754758184\n",
      "R2:  -8.249986264118114\n",
      "VOD Performance: \n",
      "MAE:  0.09984018913600672\n",
      "MSE:  0.0373749916649063\n",
      "RMSE:  0.193326127734733\n",
      "R2:  0.44039249162382144\n"
     ]
    },
    {
     "data": {
      "text/plain": "(15.512991386137019,\n 632.1912049561365,\n 25.143412754758184,\n -8.249986264118114,\n 0.09984018913600672,\n 0.0373749916649063,\n 0.193326127734733,\n 0.44039249162382144)"
     },
     "execution_count": 226,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_map_unscaled = scaler.inverse_transform(y_map)\n",
    "print(y_map_unscaled[0:5])\n",
    "\n",
    "y_pred_map_unscaled = scaler.inverse_transform(y_pred_map)\n",
    "print(y_pred_map_unscaled[0:5])\n",
    "\n",
    "print(\"Metrics for nov 1st 2020 map, unscaled:\")\n",
    "get_metrics(y_map_unscaled, y_pred_map_unscaled)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 229,
   "outputs": [
    {
     "data": {
      "text/plain": "         lat         lon   density  snowtemp  soiltemp     omega       Tbh  \\\n0  33.165960   78.563280  0.077240  0.639721  0.885132  0.000000  0.684257   \n1  50.799440  110.306015  0.143803  0.803512  0.878953  0.702479  0.731068   \n2  52.377525  101.530080  0.145976  0.734173  0.992591  0.704958  0.726729   \n3  33.250010   80.803940  0.143302  0.689974  0.950213  0.000000  0.667114   \n4  34.605970   87.245850  0.143325  0.707016  0.865505  0.196491  0.650340   \n\n        Tbv        cf         h       voc       date  snow density  \\\n0  0.696721  0.207921  0.114647  0.000000 2020-11-01    127.034106   \n1  0.726857  0.256997  0.040906  0.254843 2020-11-01    150.330988   \n2  0.720115  0.221864  0.143078  0.197100 2020-11-01    151.091610   \n3  0.702663  0.208257  0.074864  0.000000 2020-11-01    150.155632   \n4  0.699951  0.316629  0.042201  0.000000 2020-11-01    150.163865   \n\n   density_unscaled  VOD_unscaled  density_pred_unscaled  VOD_pred_unscaled  \n0        127.034106      0.000000             151.681732          -0.005569  \n1        150.330988      0.382231             166.188248           0.493318  \n2        151.091610      0.295626             158.279633           0.569852  \n3        150.155632      0.000000             152.029907          -0.008379  \n4        150.163865      0.000000             153.207581           0.001564  ",
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>lat</th>\n      <th>lon</th>\n      <th>density</th>\n      <th>snowtemp</th>\n      <th>soiltemp</th>\n      <th>omega</th>\n      <th>Tbh</th>\n      <th>Tbv</th>\n      <th>cf</th>\n      <th>h</th>\n      <th>voc</th>\n      <th>date</th>\n      <th>snow density</th>\n      <th>density_unscaled</th>\n      <th>VOD_unscaled</th>\n      <th>density_pred_unscaled</th>\n      <th>VOD_pred_unscaled</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>33.165960</td>\n      <td>78.563280</td>\n      <td>0.077240</td>\n      <td>0.639721</td>\n      <td>0.885132</td>\n      <td>0.000000</td>\n      <td>0.684257</td>\n      <td>0.696721</td>\n      <td>0.207921</td>\n      <td>0.114647</td>\n      <td>0.000000</td>\n      <td>2020-11-01</td>\n      <td>127.034106</td>\n      <td>127.034106</td>\n      <td>0.000000</td>\n      <td>151.681732</td>\n      <td>-0.005569</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>50.799440</td>\n      <td>110.306015</td>\n      <td>0.143803</td>\n      <td>0.803512</td>\n      <td>0.878953</td>\n      <td>0.702479</td>\n      <td>0.731068</td>\n      <td>0.726857</td>\n      <td>0.256997</td>\n      <td>0.040906</td>\n      <td>0.254843</td>\n      <td>2020-11-01</td>\n      <td>150.330988</td>\n      <td>150.330988</td>\n      <td>0.382231</td>\n      <td>166.188248</td>\n      <td>0.493318</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>52.377525</td>\n      <td>101.530080</td>\n      <td>0.145976</td>\n      <td>0.734173</td>\n      <td>0.992591</td>\n      <td>0.704958</td>\n      <td>0.726729</td>\n      <td>0.720115</td>\n      <td>0.221864</td>\n      <td>0.143078</td>\n      <td>0.197100</td>\n      <td>2020-11-01</td>\n      <td>151.091610</td>\n      <td>151.091610</td>\n      <td>0.295626</td>\n      <td>158.279633</td>\n      <td>0.569852</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>33.250010</td>\n      <td>80.803940</td>\n      <td>0.143302</td>\n      <td>0.689974</td>\n      <td>0.950213</td>\n      <td>0.000000</td>\n      <td>0.667114</td>\n      <td>0.702663</td>\n      <td>0.208257</td>\n      <td>0.074864</td>\n      <td>0.000000</td>\n      <td>2020-11-01</td>\n      <td>150.155632</td>\n      <td>150.155632</td>\n      <td>0.000000</td>\n      <td>152.029907</td>\n      <td>-0.008379</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>34.605970</td>\n      <td>87.245850</td>\n      <td>0.143325</td>\n      <td>0.707016</td>\n      <td>0.865505</td>\n      <td>0.196491</td>\n      <td>0.650340</td>\n      <td>0.699951</td>\n      <td>0.316629</td>\n      <td>0.042201</td>\n      <td>0.000000</td>\n      <td>2020-11-01</td>\n      <td>150.163865</td>\n      <td>150.163865</td>\n      <td>0.000000</td>\n      <td>153.207581</td>\n      <td>0.001564</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 229,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Now merge back with latlon:\n",
    "latlon1['density_unscaled'] = y_map_unscaled[:,0]\n",
    "latlon1['VOD_unscaled'] = y_map_unscaled[:,1]\n",
    "latlon1['density_pred_unscaled'] = y_pred_map_unscaled[:,0]\n",
    "latlon1['VOD_pred_unscaled'] = y_pred_map_unscaled[:,1]\n",
    "latlon1.head()"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 230,
   "outputs": [],
   "source": [
    "latlon1.to_parquet('LSTM_Map_30/LSTM_Map_30_pred.parquet')"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [],
   "source": [],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Map Best performer for VOD:"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 231,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Length of df of unique latlon paris that meet cutoff:  38849\n",
      "             lat        lon\n",
      "15565  59.918694 -121.78942\n",
      "25923  63.144420 -113.38693\n",
      "28344  63.928204  121.88278\n",
      "21260  61.636730  127.76453\n",
      "33127  65.734665 -139.34130\n",
      "Length of testdate:  58314\n",
      "           density  snowtemp  soiltemp     omega       Tbh       Tbv  \\\n",
      "30586952  0.170206  0.708772  0.845780  0.681356  0.705369  0.713866   \n",
      "30586953  0.167071  0.730467  0.868891  0.680220  0.688666  0.689544   \n",
      "30586954  0.159643  0.697934  0.845582  0.680000  0.690524  0.704247   \n",
      "30586955  0.170206  0.708772  0.857352  0.681035  0.701983  0.712883   \n",
      "30586956  0.174646  0.704454  0.869552  0.700000  0.698099  0.709491   \n",
      "\n",
      "                cf         h       voc        lat        lon       date  \n",
      "30586952  0.323790  0.090114  0.064464  65.565310 -179.95332 2020-11-01  \n",
      "30586953  0.322650  0.085397  0.004672  65.229836 -179.95332 2020-11-01  \n",
      "30586954  0.314897  0.095943  0.068366  65.734665 -179.85995 2020-11-01  \n",
      "30586955  0.336332  0.088294  0.065556  65.565310 -179.85995 2020-11-01  \n",
      "30586956  0.353848  0.088024  0.064836  65.565310 -179.76660 2020-11-01  \n",
      "Length of testdate after merge:  19681\n",
      "    density  snowtemp  soiltemp     omega       Tbh       Tbv        cf  \\\n",
      "0  0.145627  0.666977  0.985380  0.700000  0.713284  0.714260  0.283224   \n",
      "1  0.144962  0.686689  0.971066  0.700826  0.716384  0.721152  0.276523   \n",
      "2  0.147401  0.663709  0.999254  0.702586  0.708530  0.707558  0.283099   \n",
      "3  0.144691  0.694415  0.993317  0.700000  0.721169  0.716186  0.262998   \n",
      "4  0.144827  0.697224  0.993719  0.700000  0.721633  0.716051  0.258180   \n",
      "\n",
      "          h       voc        lat        lon       date  \n",
      "0  0.105285  0.068258  67.493700 -162.30809 2020-11-01  \n",
      "1  0.110016  0.077796  67.676765 -162.21474 2020-11-01  \n",
      "2  0.106186  0.064648  67.312030 -162.21474 2020-11-01  \n",
      "3  0.118349  0.078642  68.047226 -162.12137 2020-11-01  \n",
      "4  0.120372  0.078642  68.047226 -162.02802 2020-11-01  \n",
      "Final len testdate:  19681\n",
      "Final len latlon1:  19681\n"
     ]
    }
   ],
   "source": [
    "test1 = test[test['date'] <= '2020-11-1']  #All records up to our map day\n",
    "\n",
    "latlon1 = latlon_setup(test1, 80) #For VOD we use the n_input = 80 length, as this was the best performer.\n",
    "print(latlon1.head())  #Maximum possible pixels.\n",
    "\n",
    "testdate = test[test['date'] == '2020-11-1']  #our y_test, essentially.\n",
    "print(\"Length of testdate: \", len(testdate))\n",
    "print(testdate.head())  #The list of pixels present on map day\n",
    "\n",
    "#Only keep pixels on testdate that are also present in latlon1:\n",
    "testdate = pd.merge(testdate, latlon1, how='inner', on=['lat', 'lon'])\n",
    "latlon1 = pd.merge(latlon1, testdate, how='inner', on=['lat', 'lon'])\n",
    "print(\"Length of testdate after merge: \", len(testdate))\n",
    "print(testdate.head())\n",
    "print(\"Final len testdate: \", len(testdate))  #80: 19681, 30: 42596\n",
    "print(\"Final len latlon1: \", len(latlon1))"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 232,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Started at:  2022-12-21 17:53:01\n",
      "Counter:  1000 , Shape X:  (1000, 80, 7) 2022-12-21 17:53:30\n",
      "Wipeout:  0\n",
      "Counter:  2000 , Shape X:  (1000, 80, 7) 2022-12-21 17:54:02\n",
      "Wipeout:  1\n",
      "Counter:  3000 , Shape X:  (1000, 80, 7) 2022-12-21 17:54:30\n",
      "Wipeout:  2\n",
      "Counter:  4000 , Shape X:  (1000, 80, 7) 2022-12-21 17:54:57\n",
      "Wipeout:  3\n",
      "Counter:  5000 , Shape X:  (1000, 80, 7) 2022-12-21 17:55:23\n",
      "Wipeout:  4\n",
      "Counter:  6000 , Shape X:  (1000, 80, 7) 2022-12-21 17:55:50\n",
      "Wipeout:  5\n",
      "Counter:  7000 , Shape X:  (1000, 80, 7) 2022-12-21 17:56:18\n",
      "Wipeout:  6\n",
      "Counter:  8000 , Shape X:  (1000, 80, 7) 2022-12-21 17:56:47\n",
      "Wipeout:  7\n",
      "Counter:  9000 , Shape X:  (1000, 80, 7) 2022-12-21 17:57:15\n",
      "Wipeout:  8\n",
      "Counter:  10000 , Shape X:  (1000, 80, 7) 2022-12-21 17:57:41\n",
      "Wipeout:  9\n",
      "Counter:  11000 , Shape X:  (1000, 80, 7) 2022-12-21 17:58:09\n",
      "Wipeout:  10\n",
      "Counter:  12000 , Shape X:  (1000, 80, 7) 2022-12-21 17:58:38\n",
      "Wipeout:  11\n",
      "Counter:  13000 , Shape X:  (1000, 80, 7) 2022-12-21 17:59:13\n",
      "Wipeout:  12\n",
      "Counter:  14000 , Shape X:  (1000, 80, 7) 2022-12-21 17:59:48\n",
      "Wipeout:  13\n",
      "Counter:  15000 , Shape X:  (1000, 80, 7) 2022-12-21 18:00:20\n",
      "Wipeout:  14\n",
      "Counter:  16000 , Shape X:  (1000, 80, 7) 2022-12-21 18:00:55\n",
      "Wipeout:  15\n",
      "Counter:  17000 , Shape X:  (1000, 80, 7) 2022-12-21 18:01:26\n",
      "Wipeout:  16\n",
      "Counter:  18000 , Shape X:  (1000, 80, 7) 2022-12-21 18:01:58\n",
      "Wipeout:  17\n",
      "Counter:  19000 , Shape X:  (1000, 80, 7) 2022-12-21 18:02:30\n",
      "Wipeout:  18\n",
      "Save last: \n",
      "Done! And it only took from  2022-12-21 17:53:01  to  2022-12-21 18:02:53\n"
     ]
    }
   ],
   "source": [
    "n_input = 80\n",
    "n_features = 7\n",
    "X = np.empty([0, n_input, n_features])\n",
    "y = np.empty([0, 2])\n",
    "#print(X.shape)\n",
    "#print(y.shape)\n",
    "path = 'LSTM_Map_80/LSTM_Map_Data_80_'\n",
    "counter = 0\n",
    "savept = 0\n",
    "data_prep_last_only(test1, latlon1, path, n_input, n_features, X, y, counter, savept)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 233,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test:  <_io.BufferedReader name='LSTM_Map_80\\\\LSTM_Map_Data_80_0.npy'> (1000, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_80\\\\LSTM_Map_Data_80_1.npy'> (2000, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_80\\\\LSTM_Map_Data_80_10.npy'> (3000, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_80\\\\LSTM_Map_Data_80_11.npy'> (4000, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_80\\\\LSTM_Map_Data_80_12.npy'> (5000, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_80\\\\LSTM_Map_Data_80_13.npy'> (6000, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_80\\\\LSTM_Map_Data_80_14.npy'> (7000, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_80\\\\LSTM_Map_Data_80_15.npy'> (8000, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_80\\\\LSTM_Map_Data_80_16.npy'> (9000, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_80\\\\LSTM_Map_Data_80_17.npy'> (10000, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_80\\\\LSTM_Map_Data_80_18.npy'> (11000, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_80\\\\LSTM_Map_Data_80_2.npy'> (12000, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_80\\\\LSTM_Map_Data_80_3.npy'> (13000, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_80\\\\LSTM_Map_Data_80_4.npy'> (14000, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_80\\\\LSTM_Map_Data_80_5.npy'> (15000, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_80\\\\LSTM_Map_Data_80_6.npy'> (16000, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_80\\\\LSTM_Map_Data_80_7.npy'> (17000, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_80\\\\LSTM_Map_Data_80_8.npy'> (18000, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_80\\\\LSTM_Map_Data_80_9.npy'> (19000, 80, 7)\n",
      "Test:  <_io.BufferedReader name='LSTM_Map_80\\\\LSTM_Map_Data_80__last.npy'> (19681, 80, 7)\n",
      "(19681, 80, 7)\n",
      "(19681, 2)\n",
      "[[0.1448367  0.20626009]\n",
      " [0.14241871 0.14182583]\n",
      " [0.1439898  0.15054344]\n",
      " [0.1461293  0.16545983]\n",
      " [0.1435301  0.06631776]]\n"
     ]
    }
   ],
   "source": [
    "X_map2 = np.empty([0, n_input, n_features])\n",
    "y_map2 = np.empty([0, 2])\n",
    "path = 'LSTM_Map_80'\n",
    "X_map2, y_map2 = assemble_test(X_map2, y_map2, path)\n",
    "\n",
    "print(X_map2.shape)\n",
    "print(y_map2.shape)\n",
    "print(y_map2[0:5])"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 234,
   "outputs": [],
   "source": [
    "model80 = keras.models.load_model(\"ModelCheckpoints80/model80sparse_final.h5\")"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 235,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "40/40 [==============================] - 16s 366ms/step\n"
     ]
    }
   ],
   "source": [
    "y_pred_map2 = model80.predict(X_map2, batch_size=batch_size, callbacks=callbacks_list, verbose=1)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 236,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[1.50692829e+02 3.09363981e-01]\n",
      " [1.49846532e+02 2.12720757e-01]\n",
      " [1.50396414e+02 2.25796081e-01]\n",
      " [1.51145241e+02 2.48168768e-01]\n",
      " [1.50235521e+02 9.94682281e-02]]\n",
      "[[157.51984      0.29612124]\n",
      " [152.01326      0.2510319 ]\n",
      " [149.27347      0.34343573]\n",
      " [153.90227      0.27100435]\n",
      " [153.64041      0.30541503]]\n"
     ]
    }
   ],
   "source": [
    "y_map2_unscaled = scaler.inverse_transform(y_map2)\n",
    "print(y_map2_unscaled[0:5])\n",
    "\n",
    "y_pred_map2_unscaled = scaler.inverse_transform(y_pred_map2)\n",
    "print(y_pred_map2_unscaled[0:5])"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 237,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Metrics for nov 1st 2020 map, unscaled:\n",
      "Snow Density Performance: \n",
      "MAE:  19.121790100825795\n",
      "MSE:  1134.7667102886744\n",
      "RMSE:  33.68629855428872\n",
      "R2:  -15.709700858592182\n",
      "VOD Performance: \n",
      "MAE:  0.08992430075433043\n",
      "MSE:  0.028290506647678224\n",
      "RMSE:  0.16819781998491604\n",
      "R2:  0.28307782441189866\n"
     ]
    },
    {
     "data": {
      "text/plain": "(19.121790100825795,\n 1134.7667102886744,\n 33.68629855428872,\n -15.709700858592182,\n 0.08992430075433043,\n 0.028290506647678224,\n 0.16819781998491604,\n 0.28307782441189866)"
     },
     "execution_count": 237,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(\"Metrics for nov 1st 2020 map, unscaled:\")\n",
    "get_metrics(y_map2_unscaled, y_pred_map2_unscaled)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 239,
   "outputs": [],
   "source": [
    "#Now merge back with latlon:\n",
    "latlon1['density_unscaled'] = y_map2_unscaled[:, 0]\n",
    "latlon1['VOD_unscaled'] = y_map2_unscaled[:, 1]\n",
    "latlon1['density_pred_unscaled'] = y_pred_map2_unscaled[:, 0]\n",
    "latlon1['VOD_pred_unscaled'] = y_pred_map2_unscaled[:, 1]\n",
    "latlon1.head()\n",
    "latlon1.to_parquet('LSTM_Map_80/LSTM_Map_80_pred.parquet')"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [],
   "source": [],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  }
 ]
}