Part1_Statistical_Learning_And_NN.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "NXSRfleA0Ojf"
   },
   "source": [
    "# CS 410 Final Project - Text Classification Competition"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "eokmpRwk0Ojl"
   },
   "source": [
    "## Zhenchen Yu (zy23)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "lK3-PCMl0nIV"
   },
   "source": [
    "### Colab Setup\n",
    "#### Skip this part if running at local machine"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "YMDpkbCS0n0G",
    "outputId": "5cd8c1ed-26d1-4ff2-dc8d-174645a1f71c"
   },
   "outputs": [],
   "source": [
    "#from google.colab import drive\n",
    "#drive.mount('/content/gdrive')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "6KPUeaqF1JRy",
    "outputId": "b6915b40-e0fd-4829-d441-0c9b431d24df"
   },
   "outputs": [],
   "source": [
    "#Check if the mount is successful\n",
    "#import os\n",
    "#os.chdir('/content/gdrive/My Drive/Deep Learning/CS410_Final')\n",
    "#!ls"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "5DVgZEW40Ojm"
   },
   "source": [
    "# Part 1: Data Prep"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "vFZYLf8Y0Ojn"
   },
   "source": [
    "## Load Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "id": "_JJQzkbR0Ojo"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.svm import SVC\n",
    "from sklearn.metrics import f1_score\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from xgboost import XGBClassifier\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "import matplotlib.pyplot as plt\n",
    "import random\n",
    "import keras\n",
    "from keras.layers import Dense, Dropout, LSTM,  BatchNormalization\n",
    "from keras.layers import Activation, Flatten, Conv2D, MaxPooling2D, Conv1D, MaxPooling1D\n",
    "from keras.models import Sequential"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "id": "qqlCTkmF0Ojt"
   },
   "outputs": [],
   "source": [
    "document_path= 'data/train.jsonl'\n",
    "def load_dataset(document_path):\n",
    "    df = pd.read_json(document_path,lines=True)\n",
    "    sentences = df.response.values\n",
    "    labels = df.label.values \n",
    "    labels = np.where(labels=='SARCASM', 1, labels) \n",
    "    labels = np.where(labels=='NOT_SARCASM', 0, labels) \n",
    "    tokenized_texts = sentences\n",
    "    return tokenized_texts,labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "YTB6g1SJ0Ojx",
    "outputId": "bf9a7f7f-08ec-41e5-82b1-6396fb99dc25",
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\rayyu\\AppData\\Local\\Continuum\\anaconda3\\envs\\ml\\lib\\site-packages\\sklearn\\feature_extraction\\text.py:301: UserWarning: Your stop_words may be inconsistent with your preprocessing. Tokenizing the stop words generated tokens ['doesn'] not in stop_words.\n",
      "  'stop_words.' % sorted(inconsistent))\n"
     ]
    }
   ],
   "source": [
    "tokenized_texts,labels = load_dataset(document_path)\n",
    "\n",
    "#Remove stopwords\n",
    "stop_word = open(\"stopwords.txt\", \"r\")\n",
    "stop_word = stop_word.read().splitlines()    \n",
    "stop_word.append('@')\n",
    "stop_word.append('user')\n",
    "\n",
    "cv=CountVectorizer(stop_words=stop_word)\n",
    "res_mrx=cv.fit_transform(tokenized_texts)\n",
    "labels=np.asarray(labels).astype(None)\n",
    "X_train, X_test, y_train, y_test = train_test_split(res_mrx, labels, test_size=0.2, random_state=42)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Prepare Prediction Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_path = 'data/test.jsonl'\n",
    "df_test = pd.read_json(test_path,lines=True)\n",
    "test_texts = df_test.response.values\n",
    "res_test=cv.transform(test_texts)\n",
    "\n",
    "#res_test is the test dataset\n",
    "#Set cnn=true for cnn model prediction\n",
    "#set rnn=true for rnn model prediction\n",
    "def make_prediction(model, cnn=False, rnn=False):    \n",
    "    if cnn==True:\n",
    "        pred = model.predict(np.array(res_test.toarray()).reshape(res_test.shape[0],res_test.shape[1],1))\n",
    "    elif rnn==True:\n",
    "        pred = model.predict(np.array(res_test.toarray()).reshape(res_test.shape[0],1,res_test.shape[1]))\n",
    "    else:\n",
    "        pred = model.predict(res_test)\n",
    "    \n",
    "    if cnn==True or rnn==True:\n",
    "        pred = np.where(pred>0.5,1,0)\n",
    "    \n",
    "    res=[]\n",
    "    for i in range(len(pred)):\n",
    "        if pred[i] == 0:\n",
    "            res.append('NOT_SARCASM')\n",
    "        else:\n",
    "            res.append('SARCASM')\n",
    "\n",
    "    twitter_id=[]\n",
    "    for i in range(1800):\n",
    "        twitter_id.append('twitter_' + str(i+1))\n",
    "\n",
    "    results = zip(twitter_id,res)\n",
    "    res2=[]\n",
    "    for i ,j in results:\n",
    "        res2.append(str(i)+','+str(j))\n",
    " \n",
    "    with open('answer.txt', 'w') as filehandle:\n",
    "        filehandle.writelines(\"%s\\n\" % place for place in res2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "b-L9AJTY0Oj2"
   },
   "source": [
    "# Part 2: Statistical Learning Models"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "80ZhY2de0Oj2"
   },
   "source": [
    "## SVM"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "E_C8YM7E0Oj3"
   },
   "source": [
    "#### Linear Kernel"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "qT8iZsa00Oj4",
    "outputId": "d12f69e0-e51a-4e1a-b30f-18051ed95096"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1 Score for Linear SVM is 0.7601809954751132\n"
     ]
    }
   ],
   "source": [
    "C_range = [0.01, 0.1, 1, 10, 100]\n",
    "best_f1, best_c = 0, 0\n",
    "for C in C_range:\n",
    "    svm_li = SVC(kernel='linear',C=C)\n",
    "    svm_li.fit(X_train, y_train)\n",
    "    y_pred = svm_li.predict(X_test)\n",
    "    if f1_score(y_test,y_pred) > best_f1:\n",
    "        best_f1 = f1_score(y_test,y_pred)\n",
    "        best_c = C\n",
    "print(\"F1 Score for Linear SVM is\", best_f1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Generate answer.txt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "make_prediction(svm_li)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Final Prediction Result: \n",
    "Precision: 0.5800395256916996\n",
    "Recall:    0.6522222222222223\n",
    "F1 Score:  0.6140167364016738"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### RBF Kernel"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "8PPtsYSI0Oj9",
    "outputId": "55d7ccf6-ccb0-4982-f957-3bd302b4f150"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1 Score for RBF SVM is 0.7586206896551725\n"
     ]
    }
   ],
   "source": [
    "C_range = [0.01, 0.1, 1, 10, 100]\n",
    "gamma_range = [0.01, 0.1, 1, 10, 100]\n",
    "best_f1,best_c,best_gamma = 0,0,0\n",
    "for C in C_range:\n",
    "    for gamma in gamma_range:\n",
    "        svm_rbf = SVC(kernel='rbf',C=C, gamma=gamma)\n",
    "        svm_rbf.fit(X_train, y_train)\n",
    "        y_pred = svm_rbf.predict(X_test)\n",
    "        if f1_score(y_test,y_pred) > best_f1:\n",
    "            best_f1 = f1_score(y_test,y_pred)\n",
    "            best_c = C\n",
    "            best_gamma = gamma\n",
    "print(\"F1 Score for RBF SVM is\",best_f1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Generate answer.txt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "make_prediction(svm_rbf)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Final Prediction Result: \n",
    "Precision: 0.500277932184547\n",
    "Recall:    1\n",
    "F1 Score:  0.6669136717302705"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "eIiihAes0OkB"
   },
   "source": [
    "### Random Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "aNzLrgtO0OkC",
    "outputId": "13c3c7ed-528d-40ef-be2a-b288928625f7"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1 Score for Random Forest is 0.7321578505457599\n"
     ]
    }
   ],
   "source": [
    "best_f1, best_n = 0, 0\n",
    "for n_est in range(10,100,2):\n",
    "    clf_rf = RandomForestClassifier(max_depth = 3, random_state=0, n_estimators = n_est)\n",
    "    clf_rf.fit(X_train, y_train)\n",
    "    y_pred = clf_rf.predict(X_test)\n",
    "    if f1_score(y_test,y_pred) > best_f1:\n",
    "        best_f1 = f1_score(y_test,y_pred)\n",
    "        best_n = n_est\n",
    "print(\"F1 Score for Random Forest is\",best_f1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "make_prediction(clf_rf)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Final Prediction Result: \n",
    "Precision: 0.5672609400324149\n",
    "Recall:    0.7777777777777778\n",
    "F1 Score:  0.6560449859418931"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "8eF-0kLJ0OkG"
   },
   "source": [
    "### XGBoost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "BAwNsjkS0OkH",
    "outputId": "01f1703f-d035-4e6d-af92-4651227c8e09"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1 Score for XGBoost is 0.7440585009140768\n"
     ]
    }
   ],
   "source": [
    "xgboost_model = XGBClassifier()\n",
    "xgboost_model.fit(X_train, y_train)\n",
    "y_pred = xgboost_model.predict(X_test)\n",
    "best_f1 = f1_score(y_test,y_pred)\n",
    "print(\"F1 Score for XGBoost is\",best_f1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "make_prediction(xgboost_model)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Final Prediction Result: \n",
    "Precision: 0.5838198498748958\n",
    "Recall:    0.7777777777777778\n",
    "F1 Score:  0.6669842782277275"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "q1jUYFje0OkL"
   },
   "source": [
    "### Logistic Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "jGxDMb8j0OkL",
    "outputId": "4a1fad60-fed2-46f6-cc60-f65d29d4c632"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1 Score for Logistic Regression is 0.7566089334548769\n"
     ]
    }
   ],
   "source": [
    "C_range = [0.01, 0.1, 1, 10, 100]\n",
    "best_f1, best_c = 0, 0\n",
    "for C in C_range:\n",
    "    clf_lr = LogisticRegression(random_state=0, solver='liblinear', C=C).fit(X_train, y_train)\n",
    "    clf_lr.fit(X_train, y_train)\n",
    "    y_pred = clf_lr.predict(X_test)\n",
    "    if f1_score(y_test,y_pred) > best_f1:\n",
    "        best_f1 = f1_score(y_test,y_pred)\n",
    "        best_c = C\n",
    "print(\"F1 Score for Logistic Regression is\", best_f1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "make_prediction(clf_lr)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Final Prediction Result: \n",
    "Precision: 0.5896180215475024\n",
    "Recall:    0.6688888888888889\n",
    "F1 Score:  0.6267568974492451"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "ts2p2SvB0OkP"
   },
   "source": [
    "## Part 3: Deep Learning Models"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "1Yg8XBQ30OkT"
   },
   "source": [
    "### CNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "id": "tBTmZCqb0OkU"
   },
   "outputs": [],
   "source": [
    "def create_model(xtrain, ytrain):\n",
    "    model=Sequential()\n",
    "    model.add(Conv1D(32,1,input_shape=xtrain.shape[1:]))\n",
    "    model.add(Activation(\"relu\"))\n",
    "    model.add(MaxPooling1D(pool_size=3))\n",
    "    model.add(Conv1D(32,1))\n",
    "    model.add(Activation(\"relu\"))\n",
    "    model.add(MaxPooling1D(pool_size=3))\n",
    "    model.add(Flatten())\n",
    "    model.add(Dense(32))\n",
    "    model.add(Activation(\"relu\"))\n",
    "    model.add(Dense(1, activation='sigmoid'))    \n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "id": "EzJ8sCdQ0Oka"
   },
   "outputs": [],
   "source": [
    "# Parameters\n",
    "EPOCHS=50\n",
    "BATCH_SIZE=64"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "id": "EXR_klIR0Okf"
   },
   "outputs": [],
   "source": [
    "xtrain = np.array(X_train.toarray()).reshape(X_train.shape[0],X_train.shape[1],1)\n",
    "ytrain=np.array(y_train)\n",
    "xtest = np.array(X_test.toarray()).reshape(X_test.shape[0],X_test.shape[1],1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "03d1wLg90Okj",
    "outputId": "6fcac462-d3fc-4161-bfdd-a0749df026e2"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From C:\\Users\\rayyu\\AppData\\Local\\Continuum\\anaconda3\\envs\\ml\\lib\\site-packages\\tensorflow\\python\\framework\\op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Colocations handled automatically by placer.\n",
      "WARNING:tensorflow:From C:\\Users\\rayyu\\AppData\\Local\\Continuum\\anaconda3\\envs\\ml\\lib\\site-packages\\tensorflow\\python\\ops\\math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use tf.cast instead.\n",
      "Train on 4000 samples, validate on 1000 samples\n",
      "Epoch 1/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.6917 - acc: 0.5277 - val_loss: 0.6908 - val_acc: 0.5530\n",
      "Epoch 2/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.6887 - acc: 0.5790 - val_loss: 0.6883 - val_acc: 0.5650\n",
      "Epoch 3/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.6846 - acc: 0.6195 - val_loss: 0.6842 - val_acc: 0.6190\n",
      "Epoch 4/50\n",
      "4000/4000 [==============================] - 19s 5ms/step - loss: 0.6787 - acc: 0.6933 - val_loss: 0.6791 - val_acc: 0.6700\n",
      "Epoch 5/50\n",
      "4000/4000 [==============================] - 18s 4ms/step - loss: 0.6721 - acc: 0.7340 - val_loss: 0.6739 - val_acc: 0.6900\n",
      "Epoch 6/50\n",
      "4000/4000 [==============================] - 19s 5ms/step - loss: 0.6655 - acc: 0.7375 - val_loss: 0.6687 - val_acc: 0.6930\n",
      "Epoch 7/50\n",
      "4000/4000 [==============================] - 18s 4ms/step - loss: 0.6588 - acc: 0.7463 - val_loss: 0.6636 - val_acc: 0.6990\n",
      "Epoch 8/50\n",
      "4000/4000 [==============================] - 18s 4ms/step - loss: 0.6521 - acc: 0.7470 - val_loss: 0.6585 - val_acc: 0.7020\n",
      "Epoch 9/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.6456 - acc: 0.7498 - val_loss: 0.6536 - val_acc: 0.7050\n",
      "Epoch 10/50\n",
      "4000/4000 [==============================] - 18s 4ms/step - loss: 0.6393 - acc: 0.7537 - val_loss: 0.6489 - val_acc: 0.6950\n",
      "Epoch 11/50\n",
      "4000/4000 [==============================] - 18s 4ms/step - loss: 0.6330 - acc: 0.7512 - val_loss: 0.6441 - val_acc: 0.7080\n",
      "Epoch 12/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.6267 - acc: 0.7565 - val_loss: 0.6395 - val_acc: 0.7080\n",
      "Epoch 13/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.6205 - acc: 0.7570 - val_loss: 0.6350 - val_acc: 0.7140\n",
      "Epoch 14/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.6144 - acc: 0.7550 - val_loss: 0.6307 - val_acc: 0.7080\n",
      "Epoch 15/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.6083 - acc: 0.7595 - val_loss: 0.6264 - val_acc: 0.7040\n",
      "Epoch 16/50\n",
      "4000/4000 [==============================] - 18s 4ms/step - loss: 0.6027 - acc: 0.7610 - val_loss: 0.6224 - val_acc: 0.7110\n",
      "Epoch 17/50\n",
      "4000/4000 [==============================] - 18s 4ms/step - loss: 0.5969 - acc: 0.7618 - val_loss: 0.6184 - val_acc: 0.7110\n",
      "Epoch 18/50\n",
      "4000/4000 [==============================] - 18s 4ms/step - loss: 0.5914 - acc: 0.7615 - val_loss: 0.6148 - val_acc: 0.7120\n",
      "Epoch 19/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.5864 - acc: 0.7660 - val_loss: 0.6113 - val_acc: 0.7110\n",
      "Epoch 20/50\n",
      "4000/4000 [==============================] - 19s 5ms/step - loss: 0.5803 - acc: 0.7670 - val_loss: 0.6078 - val_acc: 0.7060\n",
      "Epoch 21/50\n",
      "4000/4000 [==============================] - 18s 4ms/step - loss: 0.5754 - acc: 0.7615 - val_loss: 0.6045 - val_acc: 0.7120\n",
      "Epoch 22/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.5710 - acc: 0.7652 - val_loss: 0.6012 - val_acc: 0.7090\n",
      "Epoch 23/50\n",
      "4000/4000 [==============================] - 19s 5ms/step - loss: 0.5656 - acc: 0.7640 - val_loss: 0.5980 - val_acc: 0.7120\n",
      "Epoch 24/50\n",
      "4000/4000 [==============================] - 19s 5ms/step - loss: 0.5608 - acc: 0.7710 - val_loss: 0.5952 - val_acc: 0.7170\n",
      "Epoch 25/50\n",
      "4000/4000 [==============================] - 18s 4ms/step - loss: 0.5564 - acc: 0.7670 - val_loss: 0.5923 - val_acc: 0.7160\n",
      "Epoch 26/50\n",
      "4000/4000 [==============================] - 18s 4ms/step - loss: 0.5517 - acc: 0.7712 - val_loss: 0.5898 - val_acc: 0.7170\n",
      "Epoch 27/50\n",
      "4000/4000 [==============================] - 18s 4ms/step - loss: 0.5475 - acc: 0.7698 - val_loss: 0.5872 - val_acc: 0.7160\n",
      "Epoch 28/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.5430 - acc: 0.7720 - val_loss: 0.5849 - val_acc: 0.7130\n",
      "Epoch 29/50\n",
      "4000/4000 [==============================] - 18s 4ms/step - loss: 0.5391 - acc: 0.7732 - val_loss: 0.5825 - val_acc: 0.7160\n",
      "Epoch 30/50\n",
      "4000/4000 [==============================] - 19s 5ms/step - loss: 0.5350 - acc: 0.7747 - val_loss: 0.5804 - val_acc: 0.7180\n",
      "Epoch 31/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.5311 - acc: 0.7752 - val_loss: 0.5784 - val_acc: 0.7220\n",
      "Epoch 32/50\n",
      "4000/4000 [==============================] - 20s 5ms/step - loss: 0.5273 - acc: 0.7762 - val_loss: 0.5764 - val_acc: 0.7190\n",
      "Epoch 33/50\n",
      "4000/4000 [==============================] - 20s 5ms/step - loss: 0.5239 - acc: 0.7758 - val_loss: 0.5746 - val_acc: 0.7230\n",
      "Epoch 34/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.5202 - acc: 0.7770 - val_loss: 0.5732 - val_acc: 0.7160\n",
      "Epoch 35/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.5166 - acc: 0.7777 - val_loss: 0.5712 - val_acc: 0.7200\n",
      "Epoch 36/50\n",
      "4000/4000 [==============================] - 20s 5ms/step - loss: 0.5131 - acc: 0.7775 - val_loss: 0.5696 - val_acc: 0.7250\n",
      "Epoch 37/50\n",
      "4000/4000 [==============================] - 19s 5ms/step - loss: 0.5097 - acc: 0.7765 - val_loss: 0.5689 - val_acc: 0.7210\n",
      "Epoch 38/50\n",
      "4000/4000 [==============================] - 20s 5ms/step - loss: 0.5068 - acc: 0.7878 - val_loss: 0.5672 - val_acc: 0.7220\n",
      "Epoch 39/50\n",
      "4000/4000 [==============================] - 19s 5ms/step - loss: 0.5037 - acc: 0.7800 - val_loss: 0.5655 - val_acc: 0.7230\n",
      "Epoch 40/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.5004 - acc: 0.7823 - val_loss: 0.5643 - val_acc: 0.7220\n",
      "Epoch 41/50\n",
      "4000/4000 [==============================] - 19s 5ms/step - loss: 0.4972 - acc: 0.7845 - val_loss: 0.5632 - val_acc: 0.7230\n",
      "Epoch 42/50\n",
      "4000/4000 [==============================] - 19s 5ms/step - loss: 0.4942 - acc: 0.7815 - val_loss: 0.5624 - val_acc: 0.7170\n",
      "Epoch 43/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.4914 - acc: 0.7855 - val_loss: 0.5612 - val_acc: 0.7230\n",
      "Epoch 44/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.4889 - acc: 0.7830 - val_loss: 0.5602 - val_acc: 0.7210\n",
      "Epoch 45/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.4859 - acc: 0.7840 - val_loss: 0.5597 - val_acc: 0.7230\n",
      "Epoch 46/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.4833 - acc: 0.7837 - val_loss: 0.5587 - val_acc: 0.7170\n",
      "Epoch 47/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.4805 - acc: 0.7867 - val_loss: 0.5581 - val_acc: 0.7220\n",
      "Epoch 48/50\n",
      "4000/4000 [==============================] - 18s 4ms/step - loss: 0.4781 - acc: 0.7857 - val_loss: 0.5571 - val_acc: 0.7180\n",
      "Epoch 49/50\n",
      "4000/4000 [==============================] - 18s 4ms/step - loss: 0.4754 - acc: 0.7850 - val_loss: 0.5566 - val_acc: 0.7180\n",
      "Epoch 50/50\n",
      "4000/4000 [==============================] - 18s 5ms/step - loss: 0.4728 - acc: 0.7867 - val_loss: 0.5560 - val_acc: 0.7180\n"
     ]
    }
   ],
   "source": [
    "# Train Model\n",
    "cnn_model=create_model(xtrain,ytrain)\n",
    "cnn_model.compile(loss='binary_crossentropy',\n",
    "         optimizer=keras.optimizers.Adam(lr=0.00001, epsilon=1e-06),\n",
    "         metrics=['accuracy'])\n",
    "cnn_hist=cnn_model.fit(xtrain,ytrain,batch_size=BATCH_SIZE,epochs=EPOCHS,validation_data=(xtest,y_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "id": "JlB-HOpZ0Okn"
   },
   "outputs": [],
   "source": [
    "#If using colab, set colab=True. If running at local, set colab=False\n",
    "def show_metrics(r,colab=False):\n",
    "    _,axes=plt.subplots(1,2,figsize=(15,5))\n",
    "    axes[0].plot(r.history['loss'], label='Train Loss')\n",
    "    axes[0].plot(r.history['val_loss'], label='Val Loss')\n",
    "    axes[0].legend()\n",
    "    axes[0].set_title('Loss')\n",
    "    if colab==False:\n",
    "        axes[1].plot(r.history['acc'], label='Train Acc')\n",
    "        axes[1].plot(r.history['val_acc'], label='Val Acc')\n",
    "        axes[1].legend()\n",
    "        axes[1].set_title('Acc')\n",
    "    else:\n",
    "        axes[1].plot(r.history['accuracy'], label='Train Acc')\n",
    "        axes[1].plot(r.history['val_accuracy'], label='Val Acc')\n",
    "        axes[1].legend()\n",
    "        axes[1].set_title('Accuracy')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 336
    },
    "id": "-66LHyer0Okq",
    "outputId": "3d153b64-02f7-4654-c2c2-3ae44444fee7",
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "show_metrics(cnn_hist)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "SlNyHUaX0Oks",
    "outputId": "2d54eebe-adad-4a85-f8b6-323fec45ae02"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1 Score for CNN is 0.7354596622889307\n"
     ]
    }
   ],
   "source": [
    "y_pred = cnn_model.predict(xtest)\n",
    "y_pred = np.where(y_pred>0.5,1,0)\n",
    "print(\"F1 Score for CNN is\", f1_score(y_pred,y_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "make_prediction(cnn_model,cnn=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Final Prediction Result:\n",
    "Precision: 0.6003521126760564\n",
    "Recall:    0.7577777777777778\n",
    "F1 Score:  0.6699410609037328"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "3N2Ed3ME0Okv"
   },
   "source": [
    "### RNN + LSTM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "id": "5KGsxFB2nkLR"
   },
   "outputs": [],
   "source": [
    "xtrain = np.array(X_train.toarray()).reshape(X_train.shape[0],1,X_train.shape[1])\n",
    "xtest = np.array(X_test.toarray()).reshape(X_test.shape[0],1,X_test.shape[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "id": "VHBpNFP80Okw"
   },
   "outputs": [],
   "source": [
    "def create_model_rnn(xtrain):\n",
    "    model=Sequential()\n",
    "    model.add(LSTM(256, input_shape=(xtrain.shape[1:]), return_sequences=True))\n",
    "    model.add(Dropout(0.3))\n",
    "    model.add(BatchNormalization())\n",
    "\n",
    "    model.add(LSTM(256, return_sequences=True))\n",
    "    model.add(Dropout(0.3))\n",
    "    model.add(BatchNormalization())\n",
    "\n",
    "    model.add(LSTM(256))\n",
    "    model.add(Dropout(0.3))\n",
    "    model.add(BatchNormalization())\n",
    "\n",
    "    model.add(Dense(32, activation=\"relu\"))\n",
    "    model.add(Dense(1, activation=\"sigmoid\"))\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "OfcCQv8L0Okz",
    "outputId": "82ea2cb7-e7da-4278-983c-d94e9ddbb4eb"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From C:\\Users\\rayyu\\AppData\\Local\\Continuum\\anaconda3\\envs\\ml\\lib\\site-packages\\keras\\backend\\tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n",
      "Train on 4000 samples, validate on 1000 samples\n",
      "Epoch 1/5\n",
      "4000/4000 [==============================] - 22s 6ms/step - loss: 0.7535 - acc: 0.5703 - val_loss: 0.6264 - val_acc: 0.6460\n",
      "Epoch 2/5\n",
      "4000/4000 [==============================] - 17s 4ms/step - loss: 0.6029 - acc: 0.6763 - val_loss: 0.5836 - val_acc: 0.6910\n",
      "Epoch 3/5\n",
      "4000/4000 [==============================] - 17s 4ms/step - loss: 0.4846 - acc: 0.7657 - val_loss: 0.5636 - val_acc: 0.7230\n",
      "Epoch 4/5\n",
      "4000/4000 [==============================] - 17s 4ms/step - loss: 0.3665 - acc: 0.8393 - val_loss: 0.5659 - val_acc: 0.7360\n",
      "Epoch 5/5\n",
      "4000/4000 [==============================] - 17s 4ms/step - loss: 0.2819 - acc: 0.8822 - val_loss: 0.5956 - val_acc: 0.7390\n"
     ]
    }
   ],
   "source": [
    "EPOCHS=5\n",
    "BATCH_SIZE = 64\n",
    "rnn_model=create_model_rnn(xtrain)\n",
    "rnn_model.compile(loss='binary_crossentropy',\n",
    "          optimizer=keras.optimizers.Adam(lr=1e-4, decay=1e-6),\n",
    "          metrics=['accuracy'])\n",
    "rnn_hist = rnn_model.fit(xtrain,y_train,batch_size=BATCH_SIZE,epochs=EPOCHS,validation_data=(xtest,y_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 336
    },
    "id": "XPaXuFr74aso",
    "outputId": "dc0d8b03-ca36-4984-8f9c-34e34d9fa6ea"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "show_metrics(rnn_hist,colab=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "o8MQXW530Ok4",
    "outputId": "e6d7cf05-2315-42fa-f54f-691d30f12ac8"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1 Score for RNN is 0.7502392344497608\n"
     ]
    }
   ],
   "source": [
    "y_pred = rnn_model.predict(xtest)\n",
    "y_pred = np.where(y_pred>0.5,1,0)\n",
    "print(\"F1 Score for RNN is\", f1_score(y_pred,y_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "make_prediction(rnn_model,rnn=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Final Prediction Result:\n",
    "Precision: 0.5967878275570583\n",
    "Recall:    0.7844444444444445\n",
    "F1 Score:  0.6778684589534325"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "Statistical Learning.ipynb",
   "provenance": []
  },
  "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": 1
}