Getting_started.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "flexible-fountain",
   "metadata": {},
   "source": [
    "# FoNS Datathon 2021 "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "naked-correspondence",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn import *\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "active-armor",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_descriptors = pd.read_csv(\"train_descriptors.csv\")\n",
    "train_mord3d = pd.read_csv(\"train_mord3d.csv\")\n",
    "train_morgan = pd.read_csv(\"train_morgan.csv\")\n",
    "train_rdk = pd.read_csv(\"train_rdk.csv\")\n",
    "\n",
    "train_crystals = pd.read_csv(\"train_crystals.csv\")\n",
    "train_distances = pd.read_csv(\"train_distances.csv\")\n",
    "train_centroid_distances = pd.read_csv(\"train_centroid_distances.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "ongoing-cornell",
   "metadata": {},
   "outputs": [],
   "source": [
    "test_descriptors = pd.read_csv(\"test_descriptors.csv\")\n",
    "test_mord3d = pd.read_csv(\"test_mord3d.csv\")\n",
    "test_morgan = pd.read_csv(\"test_morgan.csv\")\n",
    "test_rdk = pd.read_csv(\"test_rdk.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "spread-knowing",
   "metadata": {},
   "source": [
    "### Data pre-processing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "explicit-repair",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(13449, 984)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_descriptors_full = train_descriptors.iloc[:, 3:-2].dropna(axis= 1, how=\"any\")\n",
    "train_descriptors_full.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "systematic-johns",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(3363, 984)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_descriptors_full = test_descriptors[train_descriptors_full.columns]\n",
    "test_descriptors_full.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "intelligent-phone",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(13449, 124) (3363, 124)\n"
     ]
    }
   ],
   "source": [
    "train_PCA = decomposition.PCA(n_components=.95)\n",
    "scaler_for_PCA = preprocessing.StandardScaler()\n",
    "train_descriptors_PCA = train_PCA.fit_transform(scaler_for_PCA.fit_transform(train_descriptors_full))\n",
    "test_descriptors_PCA = train_PCA.transform(scaler_for_PCA.transform(test_descriptors_full))\n",
    "print(train_descriptors_PCA.shape, test_descriptors_PCA.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "arbitrary-consistency",
   "metadata": {},
   "source": [
    "### A (very!!) simple and bad model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "hollow-flexibility",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LinearRegression()"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "target = \"a\"\n",
    "model = linear_model.LinearRegression()\n",
    "model.fit(train_descriptors_PCA, train_crystals[target])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "sublime-audio",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "predictions = model.predict(train_descriptors_PCA)\n",
    "plt.plot(train_crystals[target], predictions, \"bo\")\n",
    "plt.xlabel(\"True values\")\n",
    "plt.ylabel(\"Predicted values\")\n",
    "plt.title(\"Predicted z-value against true values\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "informal-award",
   "metadata": {},
   "source": [
    "### Cross-validation to evaluate the model's performance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "musical-flash",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average negative MAE: -3.258\n",
      "Standard deviation: 0.05\n"
     ]
    }
   ],
   "source": [
    "cv_scores = model_selection.cross_val_score(\n",
    "    model, \n",
    "    train_descriptors_PCA, \n",
    "    train_crystals[target], \n",
    "    scoring=\"neg_mean_absolute_error\")\n",
    "\n",
    "print(\"Average negative MAE:\", round(np.mean(cv_scores), 3))\n",
    "print(\"Standard deviation:\", round(np.std(cv_scores), 3))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "based-invitation",
   "metadata": {},
   "source": [
    "### Outputting predictions for automated checking \n",
    "Link: https://github.com/stevenkbennett/fons_datathon_testing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "legislative-region",
   "metadata": {},
   "outputs": [],
   "source": [
    "predictions = model.predict(test_descriptors_PCA)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "original-smell",
   "metadata": {},
   "outputs": [],
   "source": [
    "np.savetxt(\"task_x_predictions.csv\", predictions)\n",
    "# Alternatively, if using pandas, you can use the following:\n",
    "pd.DataFrame(predictions).to_csv(\"task_x_predictions.csv\", header=False, index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "distant-museum",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "trained-moses",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}