Activity01.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "from nltk import word_tokenize\n",
    "from nltk.corpus import stopwords\n",
    "from nltk.stem import WordNetLemmatizer\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn.model_selection import train_test_split\n",
    "import nltk\n",
    "import warnings\n",
    "import string\n",
    "import re\n",
    "warnings.filterwarnings('ignore')\n",
    "from sklearn.metrics import accuracy_score, roc_curve, classification_report, confusion_matrix, precision_recall_curve, auc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "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>headline</th>\n",
       "      <th>short_description</th>\n",
       "      <th>is_political</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>27075</th>\n",
       "      <td>Even As A Teenager, Obama Knew People Like Tru...</td>\n",
       "      <td>\"Respect came from what you did and not who yo...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16255</th>\n",
       "      <td>Debbie Reynolds And Carrie Fisher Remembered A...</td>\n",
       "      <td>R2-D2 and Dan Aykroyd celebrated the iconic ac...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>64022</th>\n",
       "      <td>'Sweet Charity' for Shirley MacLaine? Not So M...</td>\n",
       "      <td>Sachi has written an extraordinarily touching ...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>41746</th>\n",
       "      <td>Bob McDonnell Wants The Supreme Court To Do So...</td>\n",
       "      <td>The former governor-turned-felon is asking Chi...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6465</th>\n",
       "      <td>The Toxic Tax Bill</td>\n",
       "      <td>Behind closed doors, Republicans in the U.S. S...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                headline  \\\n",
       "27075  Even As A Teenager, Obama Knew People Like Tru...   \n",
       "16255  Debbie Reynolds And Carrie Fisher Remembered A...   \n",
       "64022  'Sweet Charity' for Shirley MacLaine? Not So M...   \n",
       "41746  Bob McDonnell Wants The Supreme Court To Do So...   \n",
       "6465                                  The Toxic Tax Bill   \n",
       "\n",
       "                                       short_description  is_political  \n",
       "27075  \"Respect came from what you did and not who yo...             1  \n",
       "16255  R2-D2 and Dan Aykroyd celebrated the iconic ac...             0  \n",
       "64022  Sachi has written an extraordinarily touching ...             0  \n",
       "41746  The former governor-turned-felon is asking Chi...             1  \n",
       "6465   Behind closed doors, Republicans in the U.S. S...             1  "
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data = pd.read_csv('data/news_political_dataset.csv')\n",
    "data.sample(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def clf_model(model_type, X_train, y_train, X_valid):\n",
    "    model = model_type.fit(X_train,y_train)\n",
    "    predicted_labels = model.predict(X_valid)\n",
    "    predicted_probab = model.predict_proba(X_valid)[:,1]\n",
    "    return [predicted_labels,predicted_probab, model]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def model_evaluation(actual_values, predicted_values, predicted_probabilities):\n",
    "    cfn_mat = confusion_matrix(actual_values,predicted_values)\n",
    "    print(\"confusion matrix: \\n\",cfn_mat)\n",
    "    print(\"\\naccuracy: \",accuracy_score(actual_values,predicted_values))\n",
    "    print(\"\\nclassification report: \\n\", classification_report(actual_values,predicted_values))\n",
    "    fpr,tpr,threshold=roc_curve(actual_values, predicted_probabilities)\n",
    "    print ('\\nArea under ROC curve for validation set:', auc(fpr,tpr))\n",
    "    fig, ax = plt.subplots(figsize=(6,6))\n",
    "    ax.plot(fpr,tpr,label='Validation set AUC')\n",
    "    plt.xlabel('False Positive Rate')\n",
    "    plt.ylabel('True Positive Rate')\n",
    "    ax.legend(loc='best')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "lemmatizer = WordNetLemmatizer()\n",
    "stop_words = stopwords.words('english')\n",
    "stop_words = stop_words + list(string.printable)\n",
    "data['cleaned_headline_text'] = data['short_description'].apply(\\\n",
    "lambda x : ' '.join([lemmatizer.lemmatize(word.lower()) \\\n",
    "    for word in word_tokenize(re.sub(r'([^\\s\\w]|_)+', ' ', str(x))) if word.lower() not in stop_words]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "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>10</th>\n",
       "      <th>administration</th>\n",
       "      <th>already</th>\n",
       "      <th>also</th>\n",
       "      <th>always</th>\n",
       "      <th>america</th>\n",
       "      <th>american</th>\n",
       "      <th>another</th>\n",
       "      <th>around</th>\n",
       "      <th>away</th>\n",
       "      <th>...</th>\n",
       "      <th>white</th>\n",
       "      <th>win</th>\n",
       "      <th>without</th>\n",
       "      <th>woman</th>\n",
       "      <th>work</th>\n",
       "      <th>world</th>\n",
       "      <th>would</th>\n",
       "      <th>year</th>\n",
       "      <th>yet</th>\n",
       "      <th>york</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 200 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    10  administration  already  also  always  america  american  another  \\\n",
       "0  0.0             0.0      0.0   0.0     0.0      0.0       0.0      0.0   \n",
       "1  0.0             0.0      0.0   0.0     0.0      0.0       0.0      0.0   \n",
       "2  0.0             0.0      0.0   0.0     0.0      0.0       0.0      0.0   \n",
       "3  0.0             0.0      0.0   0.0     0.0      0.0       0.0      0.0   \n",
       "4  0.0             0.0      0.0   0.0     0.0      0.0       0.0      0.0   \n",
       "\n",
       "   around  away  ...  white  win  without  woman  work  world  would  year  \\\n",
       "0     0.0   0.0  ...    0.0  0.0      0.0    0.0   0.0    0.0    0.0   0.0   \n",
       "1     0.0   0.0  ...    0.0  0.0      0.0    0.0   0.0    0.0    0.0   0.0   \n",
       "2     0.0   0.0  ...    0.0  0.0      0.0    0.0   0.0    0.0    0.0   0.0   \n",
       "3     0.0   0.0  ...    0.0  0.0      0.0    0.0   0.0    0.0    0.0   0.0   \n",
       "4     0.0   0.0  ...    0.0  0.0      0.0    0.0   0.0    0.0    0.0   0.0   \n",
       "\n",
       "   yet  york  \n",
       "0  0.0   0.0  \n",
       "1  0.0   0.0  \n",
       "2  0.0   0.0  \n",
       "3  0.0   0.0  \n",
       "4  0.0   0.0  \n",
       "\n",
       "[5 rows x 200 columns]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "MAX_FEATURES = 200\n",
    "tfidf_model = TfidfVectorizer(max_features=MAX_FEATURES)\n",
    "tfidf_df = pd.DataFrame(tfidf_model.fit_transform(data['cleaned_headline_text']).todense())\n",
    "tfidf_df.columns = sorted(tfidf_model.vocabulary_)\n",
    "tfidf_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_valid, y_train, y_valid = train_test_split(tfidf_df, data['is_political'], test_size=0.2, random_state=42,stratify = data['is_political'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "confusion matrix: \n",
      " [[6827  526]\n",
      " [3420 3128]]\n",
      "\n",
      "accuracy:  0.7161355298179987\n",
      "\n",
      "classification report: \n",
      "               precision    recall  f1-score   support\n",
      "\n",
      "           0       0.67      0.93      0.78      7353\n",
      "           1       0.86      0.48      0.61      6548\n",
      "\n",
      "    accuracy                           0.72     13901\n",
      "   macro avg       0.76      0.70      0.69     13901\n",
      "weighted avg       0.76      0.72      0.70     13901\n",
      "\n",
      "\n",
      "Area under ROC curve for validation set: 0.7514346971357402\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from xgboost import XGBClassifier\n",
    "xgb_clf=XGBClassifier(n_estimators=10,learning_rate=0.05,max_depth=18,subsample=0.6,colsample_bytree= 0.6,reg_alpha= 10,seed=42)\n",
    "results = clf_model(xgb_clf, X_train, y_train, X_valid)\n",
    "model_evaluation(y_valid, results[0], results[1])\n",
    "model_xgb = results[2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "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>word</th>\n",
       "      <th>importance</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>178</th>\n",
       "      <td>trump</td>\n",
       "      <td>0.193803</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>136</th>\n",
       "      <td>president</td>\n",
       "      <td>0.104976</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>144</th>\n",
       "      <td>republican</td>\n",
       "      <td>0.082170</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>160</th>\n",
       "      <td>state</td>\n",
       "      <td>0.036298</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           word  importance\n",
       "178       trump    0.193803\n",
       "136   president    0.104976\n",
       "144  republican    0.082170\n",
       "160       state    0.036298"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "word_importances = pd.DataFrame({'word':X_train.columns,'importance':model_xgb.feature_importances_})\n",
    "word_importances.sort_values('importance', ascending = False).head(4)"
   ]
  }
 ],
 "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": "3.7.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}