4.TEXT_CLASSIFICATION/4.1.4 Linear Regression with Word2Vec.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4.1.3 Linear Regression Example with Word2Vec"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Word2Vec Feature Example"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import re\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "from bs4 import BeautifulSoup\n",
    "from nltk.corpus import stopwords\n",
    "from nltk.tokenize import word_tokenize"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "DATA_IN_PATH = './data_in/'\n",
    "DATA_OUT_PATH = './data_out/'\n",
    "TRAIN_CLEAN_DATA = 'train_clean.csv'\n",
    "\n",
    "RANDOM_SEED = 42\n",
    "TEST_SPLIT = 0.2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_data = pd.read_csv(DATA_IN_PATH + TRAIN_CLEAN_DATA)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "reviews = list(train_data['review'])\n",
    "sentiments = list(train_data['sentiment'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "sentences = []\n",
    "for review in reviews:\n",
    "    sentences.append(review.split())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_features = 300    \n",
    "min_word_count = 40   \n",
    "num_workers = 4       \n",
    "context = 10          \n",
    "downsampling = 1e-3 "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import logging\n",
    "logging.basicConfig(format='%(asctime)s : %(levelname)s : %(message)s',\\\n",
    "   level=logging.INFO)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from gensim.models import word2vec\n",
    "\n",
    "model = word2vec.Word2Vec(sentences, workers=num_workers, \\\n",
    "           size=num_features, min_count = min_word_count, \\\n",
    "            window = context, sample = downsampling)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_features(words, model, num_features):\n",
    "    feature_vector = np.zeros((num_features),dtype=np.float32)\n",
    "\n",
    "    num_words = 0\n",
    "    index2word_set = set(model.wv.index2word)\n",
    "\n",
    "    for w in words:\n",
    "        if w in index2word_set:\n",
    "            num_words += 1\n",
    "            feature_vector = np.add(feature_vector, model[w])\n",
    "\n",
    "    feature_vector = np.divide(feature_vector, num_words)\n",
    "    return feature_vector"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_dataset(reviews, model, num_features):\n",
    "    dataset = list()\n",
    "\n",
    "    for s in reviews:\n",
    "        dataset.append(get_features(s, model, num_features))\n",
    "\n",
    "    reviewFeatureVecs = np.stack(dataset)\n",
    "    \n",
    "    return reviewFeatureVecs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_data_vecs = get_dataset(sentences, model, num_features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "import numpy as np\n",
    "\n",
    "X = test_data_vecs\n",
    "y = np.array(sentiments)\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=TEST_SPLIT, random_state=RANDOM_SEED)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "\n",
    "lgs = LogisticRegression(class_weight='balanced')\n",
    "lgs.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "predicted = lgs.predict(X_test)\n",
    "from sklearn import metrics\n",
    "\n",
    "fpr, tpr, _ = metrics.roc_curve(y_test, (lgs.predict_proba(X_test)[:, 1]))\n",
    "auc = metrics.auc(fpr, tpr)\n",
    "\n",
    "print(\"------------\")\n",
    "print(\"Accuracy: %f\" % lgs.score(X_test, y_test))  #checking the accuracy\n",
    "print(\"Precision: %f\" % metrics.precision_score(y_test, predicted))\n",
    "print(\"Recall: %f\" % metrics.recall_score(y_test, predicted))\n",
    "print(\"F1-Score: %f\" % metrics.f1_score(y_test, predicted))\n",
    "print(\"AUC: %f\" % auc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "TEST_CLEAN_DATA = 'test_clean.csv'\n",
    "\n",
    "test_data = pd.read_csv(DATA_IN_PATH + TEST_CLEAN_DATA)\n",
    "\n",
    "test_review = list(test_data['review'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_data.head(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_sentences = list()\n",
    "for review in test_review:\n",
    "    test_sentences.append(review.split())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_data_vecs = get_dataset(test_sentences, model, num_features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_predicted = lgs.predict(test_data_vecs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "ids = list(test_data['id'])\n",
    "\n",
    "answer_dataset = pd.DataFrame({'id': ids, 'sentiment': test_predicted})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "if not os.path.exists(DATA_OUT_PATH):\n",
    "    os.makedirs(DATA_OUT_PATH)\n",
    "\n",
    "answer_dataset.to_csv(DATA_OUT_PATH + 'lgs_w2v_answer.csv', index=False, quoting=3)"
   ]
  }
 ],
 "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}