Quora_Insincere_Questions_Classification.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Quora Insincere Questions Classification.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/manashpratim/Quora-Insincere-Questions-Classification/blob/master/Quora_Insincere_Questions_Classification.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8i4x_29HjsHM",
        "colab_type": "code",
        "outputId": "f25989b0-57f3-4a8d-f0d7-d9d6ab41d44d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 239
        }
      },
      "source": [
        "#Downloading the Datasets\n",
        "!wget --no-check-certificate \\\n",
        "      \"https://storage.googleapis.com/kaggle-competitions-data/kaggle/10737/290346/all.zip?GoogleAccessId=web-data@kaggle-161607.iam.gserviceaccount.com&Expires=1562496110&Signature=MoI4MS2ahHrnKTYhBs6GyCrawyQLaC6muis0M%2FcaHl54q1L6RnwVwc1ElWHX3UcZQToj1YhnmIg4Nm4veYgmbE%2BGU6qwWqlejyUCrRpIZCdrvO9QoCCW52zqjxWFJpJV2QuM7pj7C7gCAjqX9CzlP0OVWesj9tGD5GSRC9wQDfWPezEXUuI08LzWOt02rbexmuNZt7axZ%2B96EKGvBhBHCzndeC75CPNYRVjVs6evEoDUyzM7%2BTtA1R8JgwQgeBJtoEdxQ15OYFaU%2FOGMLePEPzf5XkcU8tI84BFzKLlc%2FDTewheg%2F0irr0xoMx6mCLJ%2BGt%2BszJFKVKgdB6zrKuxOLw%3D%3D&response-content-disposition=attachment%3B+filename%3Dquora-insincere-questions-classification.zip\"\\\n",
        "      -O \"/tmp/quora.zip\""
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--2019-07-05 01:30:06--  https://storage.googleapis.com/kaggle-competitions-data/kaggle/10737/290346/all.zip?GoogleAccessId=web-data@kaggle-161607.iam.gserviceaccount.com&Expires=1562496110&Signature=MoI4MS2ahHrnKTYhBs6GyCrawyQLaC6muis0M%2FcaHl54q1L6RnwVwc1ElWHX3UcZQToj1YhnmIg4Nm4veYgmbE%2BGU6qwWqlejyUCrRpIZCdrvO9QoCCW52zqjxWFJpJV2QuM7pj7C7gCAjqX9CzlP0OVWesj9tGD5GSRC9wQDfWPezEXUuI08LzWOt02rbexmuNZt7axZ%2B96EKGvBhBHCzndeC75CPNYRVjVs6evEoDUyzM7%2BTtA1R8JgwQgeBJtoEdxQ15OYFaU%2FOGMLePEPzf5XkcU8tI84BFzKLlc%2FDTewheg%2F0irr0xoMx6mCLJ%2BGt%2BszJFKVKgdB6zrKuxOLw%3D%3D&response-content-disposition=attachment%3B+filename%3Dquora-insincere-questions-classification.zip\n",
            "Resolving storage.googleapis.com (storage.googleapis.com)... 172.217.214.128, 2607:f8b0:4001:c00::80\n",
            "Connecting to storage.googleapis.com (storage.googleapis.com)|172.217.214.128|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 6475626955 (6.0G) [application/zip]\n",
            "Saving to: ‘/tmp/quora.zip’\n",
            "\n",
            "/tmp/quora.zip      100%[===================>]   6.03G  72.1MB/s    in 78s     \n",
            "\n",
            "2019-07-05 01:31:25 (79.0 MB/s) - ‘/tmp/quora.zip’ saved [6475626955/6475626955]\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fT8Kswg1lXNA",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Unzipping the downloaded dataset\n",
        "import os\n",
        "import zipfile\n",
        "\n",
        "local_zip='/tmp/quora.zip'\n",
        "zip_ref=zipfile.ZipFile(local_zip,'r')\n",
        "zip_ref.extractall('/tmp/quora')\n",
        "zip_ref.close()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Uf503KX6q2oR",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Unzipping the word embeddings\n",
        "import os\n",
        "import zipfile\n",
        "\n",
        "local_zip='/tmp/quora/embeddings.zip'\n",
        "zip_ref=zipfile.ZipFile(local_zip,'r')\n",
        "zip_ref.extractall('/tmp/quora/embeddings')\n",
        "zip_ref.close()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "tlrXcgANrhJ1",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#Loading the Dataset\n",
        "import pandas as pd\n",
        "df=pd.read_csv('/tmp/quora/train.csv')\n",
        "df_test=pd.read_csv('/tmp/quora/test.csv')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "R5f-fxoWs1kK",
        "colab_type": "code",
        "outputId": "b09321c7-3a84-40b8-9029-8b41a13703b0",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 143
        }
      },
      "source": [
        "df.head(3)"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "execute_result",
          "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>qid</th>\n",
              "      <th>question_text</th>\n",
              "      <th>target</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>00002165364db923c7e6</td>\n",
              "      <td>How did Quebec nationalists see their province...</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>000032939017120e6e44</td>\n",
              "      <td>Do you have an adopted dog, how would you enco...</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>0000412ca6e4628ce2cf</td>\n",
              "      <td>Why does velocity affect time? Does velocity a...</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                    qid  ... target\n",
              "0  00002165364db923c7e6  ...      0\n",
              "1  000032939017120e6e44  ...      0\n",
              "2  0000412ca6e4628ce2cf  ...      0\n",
              "\n",
              "[3 rows x 3 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0d7TbgQvZ47S",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 143
        },
        "outputId": "b7a39f63-1351-4963-8d21-61060f70fc18"
      },
      "source": [
        "df_test.head(3)"
      ],
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "execute_result",
          "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>qid</th>\n",
              "      <th>question_text</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>0000163e3ea7c7a74cd7</td>\n",
              "      <td>Why do so many women become so rude and arroga...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>00002bd4fb5d505b9161</td>\n",
              "      <td>When should I apply for RV college of engineer...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>00007756b4a147d2b0b3</td>\n",
              "      <td>What is it really like to be a nurse practitio...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                    qid                                      question_text\n",
              "0  0000163e3ea7c7a74cd7  Why do so many women become so rude and arroga...\n",
              "1  00002bd4fb5d505b9161  When should I apply for RV college of engineer...\n",
              "2  00007756b4a147d2b0b3  What is it really like to be a nurse practitio..."
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 11
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-duWAgEwtd7v",
        "colab_type": "code",
        "outputId": "8cf493b6-c78a-4682-952b-3d1b1e24be37",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 72
        }
      },
      "source": [
        "#Class Distribution\n",
        "df['target'].value_counts()"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0    1225312\n",
              "1      80810\n",
              "Name: target, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 12
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "N-bdNTbdtLhY",
        "colab_type": "code",
        "outputId": "d857babb-257c-4932-d785-f1ea0d41a9b7",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 90
        }
      },
      "source": [
        "#Functions for Preprocessing the Dataset\n",
        "import nltk\n",
        "from nltk import word_tokenize\n",
        "nltk.download('stopwords')\n",
        "from nltk.corpus import stopwords\n",
        "nltk.download('punkt')\n",
        "stop_words = set(stopwords.words('english'))\n",
        "\n",
        "\n",
        "def mystopwords(text):\n",
        "    return ' '.join([w for w in word_tokenize(text) if not w in stop_words])\n",
        "\n",
        "import re\n",
        "def clean_text(string):\n",
        "  clean=re.sub('[\\n]',' ',string) #remove newline character\n",
        "  clean=re.sub('[^a-zA-Z]',' ',clean) #remove non alphabetic characters\n",
        "  clean=mystopwords(clean) #remove stopwords\n",
        "  return clean"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[nltk_data] Downloading package stopwords to /root/nltk_data...\n",
            "[nltk_data]   Unzipping corpora/stopwords.zip.\n",
            "[nltk_data] Downloading package punkt to /root/nltk_data...\n",
            "[nltk_data]   Unzipping tokenizers/punkt.zip.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4BWdGI8QtTlB",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#Cleaning the data and separating the texts and labels\n",
        "sentences=[]\n",
        "labels=[]\n",
        "for i in range(len(df)):\n",
        "  labels.append(df['target'][i])\n",
        "  sentences.append(clean_text(df['question_text'][i]))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pnx8KTnednfr",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "test_sentences=[]\n",
        "for i in range(len(df_test)):\n",
        "  test_sentences.append(clean_text(df['question_text'][i]))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rLMGL77muVlV",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#Suffling the data and splitting the data into train and test sets (90:10)\n",
        "from sklearn.model_selection import train_test_split\n",
        "train_sentences, val_sentences, train_labels, val_labels= train_test_split(sentences, labels, test_size=0.1, shuffle= True)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pB1h2-DVu0TN",
        "colab_type": "code",
        "outputId": "9e04ee28-a6a6-4420-8ecd-9d86145401c1",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 90
        }
      },
      "source": [
        "#Data Summary\n",
        "print('Length of the Data: ',len(df))\n",
        "print('No. of Training Examples: ',len(train_sentences))\n",
        "print('No. of Validation Examples: ',len(val_sentences))\n",
        "print('No. of Classes: ',df['target'].nunique())"
      ],
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Length of the Data:  1306122\n",
            "No. of Training Examples:  1175509\n",
            "No. of Validation Examples:  130613\n",
            "No. of Classes:  2\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "f085D4OA0O4y",
        "colab_type": "text"
      },
      "source": [
        "# **Bag of Words Models**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EDyEl0KDPsTL",
        "colab_type": "text"
      },
      "source": [
        "## **Multinomial Naive Bayes Classifier**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kyuvIfWDvye8",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from sklearn.feature_extraction.text import CountVectorizer\n",
        "from sklearn.feature_extraction.text import TfidfTransformer\n",
        "from sklearn.naive_bayes import MultinomialNB\n",
        "from sklearn.pipeline import Pipeline\n",
        "pipeline=Pipeline([('bow',CountVectorizer(analyzer=clean_text)),\n",
        "                   ('tfidf',TfidfTransformer()),\n",
        "                   ('classifier',MultinomialNB())\n",
        "                   ])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mLYm-R4dP43R",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 651
        },
        "outputId": "b35c9846-e92d-40cf-af77-b862e5dd14fb"
      },
      "source": [
        "pipeline.fit(train_sentences,train_labels)\n",
        "pred=pipeline.predict(val_sentences)\n",
        "\n",
        "from sklearn.metrics import accuracy_score\n",
        "print('Test Accuracy using MultiNomial Naive Bayes Classifier: ',accuracy_score(val_labels,pred))\n",
        "\n",
        "from sklearn.metrics import classification_report\n",
        "print(classification_report(val_labels,pred))\n",
        "import matplotlib.pyplot as plt\n",
        "from sklearn.metrics import confusion_matrix\n",
        "print(confusion_matrix(val_labels, pred))\n",
        "CM = confusion_matrix(val_labels, pred)\n",
        "from mlxtend.plotting import plot_confusion_matrix\n",
        "fig, ax = plot_confusion_matrix(conf_mat=CM ,  figsize=(5, 5))\n",
        "plt.show()"
      ],
      "execution_count": 121,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/sklearn/utils/validation.py:724: DataConversionWarning: A column-vector y was passed when a 1d array was expected. Please change the shape of y to (n_samples, ), for example using ravel().\n",
            "  y = column_or_1d(y, warn=True)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Test Accuracy using MultiNomial Naive Bayes Classifier:  0.9381455138462481\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.94      1.00      0.97    122534\n",
            "           1       0.00      0.00      0.00      8079\n",
            "\n",
            "    accuracy                           0.94    130613\n",
            "   macro avg       0.47      0.50      0.48    130613\n",
            "weighted avg       0.88      0.94      0.91    130613\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/sklearn/metrics/classification.py:1437: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples.\n",
            "  'precision', 'predicted', average, warn_for)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "[[122534      0]\n",
            " [  8079      0]]\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAT0AAAFACAYAAADOJ6uCAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nWIx7PtyW2AAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 360x360 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0A87pt-lU-lq",
        "colab_type": "text"
      },
      "source": [
        "### **While the classifier has a high accuracy, it fails to classify  even a single example from the minority class **"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "38d7tgwpUdf7",
        "colab_type": "text"
      },
      "source": [
        "## **Logistic Regression**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "x0S-nppyUmVM",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from sklearn.feature_extraction.text import CountVectorizer\n",
        "from sklearn.feature_extraction.text import TfidfTransformer\n",
        "from sklearn.naive_bayes import MultinomialNB\n",
        "from sklearn.pipeline import Pipeline\n",
        "from sklearn.linear_model import LogisticRegression\n",
        "pipeline=Pipeline([('bow',CountVectorizer(analyzer=clean_text)),\n",
        "                   ('tfidf',TfidfTransformer()),\n",
        "                   ('classifier',LogisticRegression())\n",
        "                   ])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6L5z57cOU3xP",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 614
        },
        "outputId": "f726ebef-49c2-4759-e587-f202ac9d57b8"
      },
      "source": [
        "pipeline.fit(train_sentences,train_labels)\n",
        "pred=pipeline.predict(val_sentences)\n",
        "\n",
        "from sklearn.metrics import accuracy_score\n",
        "print('Test Accuracy using Logistic Regression Classifier: ',accuracy_score(val_labels,pred))\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "from sklearn.metrics import classification_report\n",
        "print(classification_report(val_labels,pred))\n",
        "\n",
        "from sklearn.metrics import confusion_matrix\n",
        "print(confusion_matrix(val_labels, pred))\n",
        "CM = confusion_matrix(val_labels, pred)\n",
        "from mlxtend.plotting import plot_confusion_matrix\n",
        "fig, ax = plot_confusion_matrix(conf_mat=CM ,  figsize=(5, 5))\n",
        "plt.show()"
      ],
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
            "  FutureWarning)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Test Accuracy using Logistic Regression Classifier:  0.938130201434773\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.94      1.00      0.97    122534\n",
            "           1       0.48      0.00      0.00      8079\n",
            "\n",
            "    accuracy                           0.94    130613\n",
            "   macro avg       0.71      0.50      0.49    130613\n",
            "weighted avg       0.91      0.94      0.91    130613\n",
            "\n",
            "[[122512     22]\n",
            " [  8059     20]]\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAT0AAAFACAYAAADOJ6uCAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nMXqSsmL0JGXF6EnKyv8BgdfwANLL0b4AAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 360x360 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vK77SvjqptcY",
        "colab_type": "text"
      },
      "source": [
        "### **This classifier could classify a few instances of the minority class. Yet, the results are abysmal.**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "RF3O5yRaYWsJ",
        "colab_type": "text"
      },
      "source": [
        "# **Word Embedding Models**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Q_tJ81sOYeqS",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        },
        "outputId": "905e4210-4a12-4208-b5d9-97c2f2919cb1"
      },
      "source": [
        "#Tokenization and Padding\n",
        "import tensorflow as tf\n",
        "import numpy as np\n",
        "\n",
        "from tensorflow.keras.preprocessing.text import Tokenizer\n",
        "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
        "\n",
        "\n",
        "embedding_dim = 300\n",
        "max_length = 100\n",
        "trunc_type='post'\n",
        "padding_type='post'\n",
        "oov_tok = \"<OOV>\"\n",
        "vocab_len=49999\n",
        "\n",
        "\n",
        "tokenizer = Tokenizer(num_words=vocab_len+1,oov_token=oov_tok)\n",
        "tokenizer.fit_on_texts(sentences)\n",
        "\n",
        "word_index = tokenizer.word_index\n",
        "vocab_size=len(word_index)\n",
        "print('Original Size of Vocabulary: ',vocab_size)\n",
        "\n",
        "word_index = {e:i for e,i in word_index.items() if i <= vocab_len+1} #Reducing the Size of Vocabulary\n",
        "print('New Size of Vocabulary: ',len(word_index))\n",
        "\n",
        "sequences = tokenizer.texts_to_sequences(sentences)\n",
        "padded = pad_sequences(sequences, maxlen=max_length, padding=padding_type, truncating=trunc_type)\n",
        "\n",
        "train_sequences = tokenizer.texts_to_sequences(train_sentences)\n",
        "train_padded = pad_sequences(train_sequences, maxlen=max_length, padding=padding_type, truncating=trunc_type)\n",
        "\n",
        "val_sequences = tokenizer.texts_to_sequences(val_sentences)\n",
        "val_padded = pad_sequences(val_sequences, maxlen=max_length, padding=padding_type, truncating=trunc_type)\n",
        "\n",
        "test_sequences = tokenizer.texts_to_sequences(test_sentences)\n",
        "test_padded = pad_sequences(test_sequences, maxlen=max_length, padding=padding_type, truncating=trunc_type)\n",
        "\n",
        "train_labels=np.expand_dims(train_labels, axis=1)\n",
        "val_labels=np.expand_dims(val_labels, axis=1)\n"
      ],
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Original Size of Vocabulary:  178863\n",
            "New Size of Vocabulary:  50000\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VYpxyjtjhEzh",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 109
        },
        "outputId": "8581ac9d-a3b5-4e7e-a610-0301fcf83e11"
      },
      "source": [
        "#Dimension of Data after Tokenizing and Padding\n",
        "print('Dimension of Training Data: ',train_padded.shape)\n",
        "print('Dimension of Validation Data: ',val_padded.shape)\n",
        "print('Dimension of Training Labels: ',train_labels.shape)\n",
        "print('Dimension of Validation Labels: ',val_labels.shape)\n",
        "print('Dimension of Test Data: ',test_padded.shape)"
      ],
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Dimension of Training Data:  (1175509, 100)\n",
            "Dimension of Validation Data:  (130613, 100)\n",
            "Dimension of Training Labels:  (1175509, 1)\n",
            "Dimension of Validation Labels:  (130613, 1)\n",
            "Dimension of Test Data:  (375806, 100)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "RBpItmz724Fq",
        "colab_type": "text"
      },
      "source": [
        "## **Bidirectional LSTM Model**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QEeuA_LKiJYf",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#Building the Bidirectional LSTM Model\n",
        "model = tf.keras.Sequential([\n",
        "    tf.keras.layers.Embedding(vocab_len+1, embedding_dim, input_length=max_length),\n",
        "    tf.keras.layers.Bidirectional(tf.keras.layers.CuDNNLSTM(64,return_sequences=True)),\n",
        "    tf.keras.layers.Flatten(),\n",
        "    tf.keras.layers.Dense(16, activation='relu'),\n",
        "    tf.keras.layers.Dropout(0.1),\n",
        "    tf.keras.layers.Dense(1, activation='sigmoid')\n",
        "])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-jVsNlOKiUxU",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 219
        },
        "outputId": "3fb976d8-92e6-44b1-b611-d5893849b420"
      },
      "source": [
        "reduce =tf. keras.callbacks.ReduceLROnPlateau(monitor='val_loss', factor=0.1, patience=3, mode='auto')\n",
        "#early = tf.keras.callbacks.EarlyStopping(monitor='val_loss', min_delta=1e-4, patience=10, mode='auto')\n",
        "model.compile(loss='binary_crossentropy',optimizer='rmsprop',metrics=['accuracy'])\n",
        "num_epochs = 5\n",
        "history=model.fit(train_padded, train_labels, epochs=num_epochs, batch_size=512, validation_data=(val_padded,val_labels),callbacks=[reduce],verbose=1)"
      ],
      "execution_count": 64,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 1175509 samples, validate on 130613 samples\n",
            "Epoch 1/5\n",
            "1175509/1175509 [==============================] - 102s 87us/sample - loss: 0.1277 - acc: 0.9514 - val_loss: 0.1168 - val_acc: 0.9544\n",
            "Epoch 2/5\n",
            "1175509/1175509 [==============================] - 101s 86us/sample - loss: 0.1158 - acc: 0.9557 - val_loss: 0.1181 - val_acc: 0.9548\n",
            "Epoch 3/5\n",
            "1175509/1175509 [==============================] - 100s 85us/sample - loss: 0.1104 - acc: 0.9580 - val_loss: 0.1129 - val_acc: 0.9562\n",
            "Epoch 4/5\n",
            "1175509/1175509 [==============================] - 101s 86us/sample - loss: 0.1060 - acc: 0.9600 - val_loss: 0.1126 - val_acc: 0.9563\n",
            "Epoch 5/5\n",
            "1175509/1175509 [==============================] - 102s 86us/sample - loss: 0.1018 - acc: 0.9618 - val_loss: 0.1169 - val_acc: 0.9535\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "j1G6_oKKlaaJ",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 201
        },
        "outputId": "a94ea538-9bd6-4712-f305-7c2d527b8f2e"
      },
      "source": [
        "#Determining the optimal threshold value to seggregate the classes\n",
        "pred1 = model.predict(val_padded, batch_size=512,verbose=1)\n",
        "from sklearn import metrics\n",
        "for thresh in np.arange(0.1, 0.501, 0.05):\n",
        "    thresh = np.round(thresh, 2)\n",
        "    print(\"F1 score at threshold {0} is {1}\".format(thresh, metrics.f1_score(val_labels, (pred1>thresh).astype(int))))"
      ],
      "execution_count": 93,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "130613/130613 [==============================] - 4s 33us/sample\n",
            "F1 score at threshold 0.1 is 0.5061702277695508\n",
            "F1 score at threshold 0.15 is 0.554305008441193\n",
            "F1 score at threshold 0.2 is 0.5837030529321326\n",
            "F1 score at threshold 0.25 is 0.6046665095451332\n",
            "F1 score at threshold 0.3 is 0.6197676747446424\n",
            "F1 score at threshold 0.35 is 0.6268546774702726\n",
            "F1 score at threshold 0.4 is 0.6338253994581743\n",
            "F1 score at threshold 0.45 is 0.6337195828505215\n",
            "F1 score at threshold 0.5 is 0.6313937261088526\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TyNBkAX3kWCx",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 539
        },
        "outputId": "2b62f10e-25c4-4e2d-ab14-18c0efc86adc"
      },
      "source": [
        "# From above, we can see that F1 score is highest at a threshold of 0.4\n",
        "pred_val_labels=(pred1>0.4).astype(int)\n",
        "\n",
        "from sklearn.metrics import classification_report\n",
        "print(classification_report(val_labels,pred_val_labels))\n",
        "\n",
        "from sklearn.metrics import confusion_matrix\n",
        "print(confusion_matrix(val_labels, pred_val_labels))\n",
        "CM = confusion_matrix(val_labels, pred_val_labels)\n",
        "from mlxtend.plotting import plot_confusion_matrix\n",
        "fig, ax = plot_confusion_matrix(conf_mat=CM ,  figsize=(5, 5))\n",
        "plt.show()"
      ],
      "execution_count": 96,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.98      0.97      0.97    122534\n",
            "           1       0.57      0.71      0.63      8079\n",
            "\n",
            "    accuracy                           0.95    130613\n",
            "   macro avg       0.78      0.84      0.80    130613\n",
            "weighted avg       0.96      0.95      0.95    130613\n",
            "\n",
            "[[118258   4276]\n",
            " [  2347   5732]]\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAT0AAAFACAYAAADOJ6uCAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 360x360 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3dS1MCn2PhGr",
        "colab_type": "text"
      },
      "source": [
        "## **Bidirectional GRU**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Z6IFv6UvO6ai",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#Building the Bidirectional GRU Model\n",
        "model_gru = tf.keras.Sequential([\n",
        "    tf.keras.layers.Embedding(vocab_len+1, embedding_dim, input_length=max_length),\n",
        "    tf.keras.layers.Bidirectional(tf.keras.layers.CuDNNGRU(64,return_sequences=True)),\n",
        "    tf.keras.layers.Flatten(),\n",
        "    tf.keras.layers.Dense(16, activation='relu'),\n",
        "    tf.keras.layers.Dropout(0.1),\n",
        "    tf.keras.layers.Dense(1, activation='sigmoid')\n",
        "])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "FycKoJRrO6w7",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 219
        },
        "outputId": "b9d5ef13-bea4-4c20-e060-57c0712a5a2b"
      },
      "source": [
        "reduce =tf. keras.callbacks.ReduceLROnPlateau(monitor='val_loss', factor=0.1, patience=3, mode='auto')\n",
        "#early = tf.keras.callbacks.EarlyStopping(monitor='val_loss', min_delta=1e-4, patience=10, mode='auto')\n",
        "model_gru.compile(loss='binary_crossentropy',optimizer='rmsprop',metrics=['accuracy'])\n",
        "num_epochs = 5\n",
        "history_gru=model_gru.fit(train_padded, train_labels, epochs=num_epochs, batch_size=512, validation_data=(val_padded,val_labels),callbacks=[reduce],verbose=1)"
      ],
      "execution_count": 74,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 1175509 samples, validate on 130613 samples\n",
            "Epoch 1/5\n",
            "1175509/1175509 [==============================] - 94s 80us/sample - loss: 0.1276 - acc: 0.9513 - val_loss: 0.1179 - val_acc: 0.9545\n",
            "Epoch 2/5\n",
            "1175509/1175509 [==============================] - 91s 78us/sample - loss: 0.1145 - acc: 0.9562 - val_loss: 0.1139 - val_acc: 0.9559\n",
            "Epoch 3/5\n",
            "1175509/1175509 [==============================] - 91s 77us/sample - loss: 0.1088 - acc: 0.9588 - val_loss: 0.1134 - val_acc: 0.9566\n",
            "Epoch 4/5\n",
            "1175509/1175509 [==============================] - 91s 77us/sample - loss: 0.1040 - acc: 0.9606 - val_loss: 0.1133 - val_acc: 0.9560\n",
            "Epoch 5/5\n",
            "1175509/1175509 [==============================] - 91s 78us/sample - loss: 0.0987 - acc: 0.9630 - val_loss: 0.1150 - val_acc: 0.9557\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ydxR3Hk5O7Og",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 182
        },
        "outputId": "cce3ceaa-7655-4572-e8a9-5b6ca5a3b217"
      },
      "source": [
        "#Determining the optimal threshold value to seggregate the classes\n",
        "pred2 = model_gru.predict(val_padded, batch_size=512,verbose=1)\n",
        "from sklearn import metrics\n",
        "for thresh in np.arange(0.1, 0.5, 0.05):\n",
        "    thresh = np.round(thresh, 2)\n",
        "    print(\"F1 score at threshold {0} is {1}\".format(thresh, metrics.f1_score(val_labels, (pred2>thresh).astype(int))))"
      ],
      "execution_count": 97,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "130613/130613 [==============================] - 4s 28us/sample\n",
            "F1 score at threshold 0.1 is 0.5567750622220409\n",
            "F1 score at threshold 0.15 is 0.5883733525907203\n",
            "F1 score at threshold 0.2 is 0.6078890841632494\n",
            "F1 score at threshold 0.25 is 0.6221012406428309\n",
            "F1 score at threshold 0.3 is 0.6281488072068064\n",
            "F1 score at threshold 0.35 is 0.6287754624209788\n",
            "F1 score at threshold 0.4 is 0.6258855417975728\n",
            "F1 score at threshold 0.45 is 0.6158066623122141\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8bId1CBrO7n6",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 539
        },
        "outputId": "e0f4b7cb-3c1e-4a83-c240-33dcee8d4a90"
      },
      "source": [
        "# From above, we can see that F1 score is highest at a threshold of 0.3\n",
        "pred_val_labels_gru=(pred2>0.3).astype(int)\n",
        "\n",
        "from sklearn.metrics import classification_report\n",
        "print(classification_report(val_labels,pred_val_labels_gru))\n",
        "\n",
        "from sklearn.metrics import confusion_matrix\n",
        "print(confusion_matrix(val_labels, pred_val_labels_gru))\n",
        "CM = confusion_matrix(val_labels, pred_val_labels_gru)\n",
        "from mlxtend.plotting import plot_confusion_matrix\n",
        "fig, ax = plot_confusion_matrix(conf_mat=CM ,  figsize=(5, 5))\n",
        "plt.show()"
      ],
      "execution_count": 100,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.98      0.97      0.97    122534\n",
            "           1       0.57      0.70      0.63      8079\n",
            "\n",
            "    accuracy                           0.95    130613\n",
            "   macro avg       0.78      0.83      0.80    130613\n",
            "weighted avg       0.95      0.95      0.95    130613\n",
            "\n",
            "[[118278   4256]\n",
            " [  2431   5648]]\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAT0AAAFACAYAAADOJ6uCAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nrBg9SVkxepKyYvQkZcXoScrK/wFqP1Xjx3KIugAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 360x360 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FbDYQOgo3aog",
        "colab_type": "text"
      },
      "source": [
        "# **Pretrained Word Embeddings**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "z-KlJRAc35zm",
        "colab_type": "text"
      },
      "source": [
        "## **fastText**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jN1CiinwvrHl",
        "colab_type": "text"
      },
      "source": [
        "**crawl-300d-2M**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3TbJmJyHurLx",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 129
        },
        "outputId": "721f0152-5712-4828-ac10-9215574d6e67"
      },
      "source": [
        "# Run this cell to mount your Google Drive.\n",
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Go to this URL in a browser: https://accounts.google.com/o/oauth2/auth?client_id=947318989803-6bn6qk8qdgf4n4g3pfee6491hc0brc4i.apps.googleusercontent.com&redirect_uri=urn%3Aietf%3Awg%3Aoauth%3A2.0%3Aoob&scope=email%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdocs.test%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdrive%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdrive.photos.readonly%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fpeopleapi.readonly&response_type=code\n",
            "\n",
            "Enter your authorization code:\n",
            "··········\n",
            "Mounted at /content/drive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QVPsvLYjvT8y",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Unzipping the fasttext pretrained model from Google Drive\n",
        "import os\n",
        "import zipfile\n",
        "\n",
        "local_zip='/content/drive/My Drive/crawl-300d-2M.vec.zip'\n",
        "zip_ref=zipfile.ZipFile(local_zip,'r')\n",
        "zip_ref.extractall('/content/drive/My Drive/crawl-300d-2M.vec')\n",
        "zip_ref.close()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "78vWlU-WvXzj",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Loading the pretrained vectors\n",
        "embeddings_index = {}\n",
        "import numpy as np\n",
        "with open('/content/drive/My Drive/crawl-300d-2M.vec/crawl-300d-2M.vec') as f:\n",
        "    for line in f:\n",
        "        values = line.split()\n",
        "        word = values[0]\n",
        "        coefs = np.asarray(values[1:], dtype='float32')\n",
        "        embeddings_index[word] = coefs"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dr9wE12mv1mi",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Creating an embedding matrix that contains the vectors of the words in the dataset\n",
        "embeddings_matrix = np.zeros((vocab_len+2, embedding_dim))\n",
        "for word, i in word_index.items():\n",
        "    embedding_vector = embeddings_index.get(word)\n",
        "    if embedding_vector is not None:\n",
        "        embeddings_matrix[i] = embedding_vector"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "b58QW2e_2JZI",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "embeddings_matrix=np.delete(embeddings_matrix, 1, 0)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bFkkEH_9wBFD",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "b4609a43-f38a-4c76-a0b8-817899be99b1"
      },
      "source": [
        "print('Shape of the Embeddings Matrix: ',embeddings_matrix.shape)"
      ],
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Shape of the Embeddings Matrix:  (50000, 300)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ASNWoba-rkmy",
        "colab_type": "text"
      },
      "source": [
        "## **Bidirectional LSTM**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hyx1B5fM2Mks",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model3_lstm = tf.keras.Sequential([\n",
        "    tf.keras.layers.Embedding(vocab_len+1, embedding_dim, input_length=max_length, weights = [embeddings_matrix], trainable = False),\n",
        "    tf.keras.layers.Bidirectional(tf.keras.layers.CuDNNLSTM(64)),\n",
        "    tf.keras.layers.Dense(16, activation='relu'),\n",
        "    tf.keras.layers.Dropout(0.1),\n",
        "    tf.keras.layers.Dense(1, activation='sigmoid')\n",
        "])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6LE67TO92Yha",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 219
        },
        "outputId": "d1847373-82d8-4850-fafe-231a59247e1b"
      },
      "source": [
        "reduce =tf. keras.callbacks.ReduceLROnPlateau(monitor='val_loss', factor=0.1, patience=3, mode='auto')\n",
        "#early = tf.keras.callbacks.EarlyStopping(monitor='val_loss', min_delta=1e-4, patience=10, mode='auto')\n",
        "model3_lstm.compile(loss='binary_crossentropy',optimizer='rmsprop',metrics=['accuracy'])\n",
        "num_epochs = 5\n",
        "history3=model3_lstm.fit(train_padded, \n",
        "                    train_labels, \n",
        "                    epochs=num_epochs, \n",
        "                    batch_size=512, \n",
        "                    validation_data=(val_padded,val_labels),\n",
        "                    callbacks=[reduce],\n",
        "                    verbose=1)"
      ],
      "execution_count": 87,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 1175509 samples, validate on 130613 samples\n",
            "Epoch 1/5\n",
            "1175509/1175509 [==============================] - 80s 68us/sample - loss: 0.1546 - acc: 0.9454 - val_loss: 0.1564 - val_acc: 0.9460\n",
            "Epoch 2/5\n",
            "1175509/1175509 [==============================] - 78s 66us/sample - loss: 0.1282 - acc: 0.9511 - val_loss: 0.1293 - val_acc: 0.9521\n",
            "Epoch 3/5\n",
            "1175509/1175509 [==============================] - 78s 66us/sample - loss: 0.1197 - acc: 0.9533 - val_loss: 0.1199 - val_acc: 0.9533\n",
            "Epoch 4/5\n",
            "1175509/1175509 [==============================] - 78s 66us/sample - loss: 0.1144 - acc: 0.9547 - val_loss: 0.1187 - val_acc: 0.9538\n",
            "Epoch 5/5\n",
            "1175509/1175509 [==============================] - 78s 66us/sample - loss: 0.1100 - acc: 0.9561 - val_loss: 0.1171 - val_acc: 0.9542\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eP_kgM7M2fa4",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 182
        },
        "outputId": "027f2147-c02d-4e32-9a4e-839aa3e488e6"
      },
      "source": [
        "#Determining the optimal threshold value to seggregate the classes\n",
        "pred3= model3_lstm.predict(val_padded, batch_size=512,verbose=1)\n",
        "from sklearn import metrics\n",
        "for thresh in np.arange(0.1, 0.5, 0.05):\n",
        "    thresh = np.round(thresh, 2)\n",
        "    print(\"F1 score at threshold {0} is {1}\".format(thresh, metrics.f1_score(val_labels, (pred3>thresh).astype(int))))"
      ],
      "execution_count": 101,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "130613/130613 [==============================] - 4s 29us/sample\n",
            "F1 score at threshold 0.1 is 0.5476286660741697\n",
            "F1 score at threshold 0.15 is 0.5804730249174694\n",
            "F1 score at threshold 0.2 is 0.6031307095945879\n",
            "F1 score at threshold 0.25 is 0.613206046141607\n",
            "F1 score at threshold 0.3 is 0.6179966044142615\n",
            "F1 score at threshold 0.35 is 0.6169499640890591\n",
            "F1 score at threshold 0.4 is 0.6103986354160086\n",
            "F1 score at threshold 0.45 is 0.6004781511488909\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_s_XRmXr2st_",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 539
        },
        "outputId": "18097cc6-9f74-4d4f-dd22-3dfe74552f8e"
      },
      "source": [
        "#From above, we can see that F1 score is highest at a threshold of 0.3\n",
        "pred_ft3_val_labels=(pred3>0.3).astype(int)\n",
        "\n",
        "from sklearn.metrics import classification_report\n",
        "print(classification_report(val_labels,pred_ft3_val_labels))\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "from sklearn.metrics import confusion_matrix\n",
        "print(confusion_matrix(val_labels, pred_ft3_val_labels))\n",
        "CM = confusion_matrix(val_labels, pred_ft3_val_labels)\n",
        "from mlxtend.plotting import plot_confusion_matrix\n",
        "fig, ax = plot_confusion_matrix(conf_mat=CM ,  figsize=(5, 5))\n",
        "plt.show()"
      ],
      "execution_count": 102,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.98      0.97      0.97    122534\n",
            "           1       0.57      0.68      0.62      8079\n",
            "\n",
            "    accuracy                           0.95    130613\n",
            "   macro avg       0.77      0.82      0.80    130613\n",
            "weighted avg       0.95      0.95      0.95    130613\n",
            "\n",
            "[[118403   4131]\n",
            " [  2619   5460]]\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAT0AAAFACAYAAADOJ6uCAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nI6AAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 360x360 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "21xw8Ld6rylt",
        "colab_type": "text"
      },
      "source": [
        "## **Bidirectional GRU**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bwpCt9BeB_QC",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model4 = tf.keras.Sequential([\n",
        "    tf.keras.layers.Embedding(vocab_len+1, embedding_dim, input_length=max_length, weights = [embeddings_matrix], trainable = False),\n",
        "    tf.keras.layers.Bidirectional(tf.keras.layers.CuDNNGRU(64)),\n",
        "    tf.keras.layers.Dense(16, activation='relu'),\n",
        "    tf.keras.layers.Dropout(0.1),\n",
        "    tf.keras.layers.Dense(1, activation='sigmoid')\n",
        "])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jx9jmqK7CEx0",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 219
        },
        "outputId": "4dae94df-0d29-4d8b-da85-fd833b15cdf3"
      },
      "source": [
        "reduce =tf. keras.callbacks.ReduceLROnPlateau(monitor='val_loss', factor=0.1, patience=10, mode='auto')\n",
        "#early = tf.keras.callbacks.EarlyStopping(monitor='val_loss', min_delta=1e-4, patience=10, mode='auto')\n",
        "model4.compile(loss='binary_crossentropy',optimizer='rmsprop',metrics=['accuracy'])\n",
        "num_epochs = 5\n",
        "history4=model4.fit(train_padded, \n",
        "                    train_labels, \n",
        "                    epochs=num_epochs, \n",
        "                    batch_size=512, \n",
        "                    validation_data=(val_padded,val_labels),\n",
        "                    callbacks=[reduce],\n",
        "                    verbose=1)"
      ],
      "execution_count": 43,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 1175509 samples, validate on 130613 samples\n",
            "Epoch 1/5\n",
            "1175509/1175509 [==============================] - 68s 58us/sample - loss: 0.1542 - acc: 0.9459 - val_loss: 0.1382 - val_acc: 0.9492\n",
            "Epoch 2/5\n",
            "1175509/1175509 [==============================] - 66s 57us/sample - loss: 0.1283 - acc: 0.9515 - val_loss: 0.1248 - val_acc: 0.9522\n",
            "Epoch 3/5\n",
            "1175509/1175509 [==============================] - 66s 56us/sample - loss: 0.1207 - acc: 0.9534 - val_loss: 0.1207 - val_acc: 0.9534\n",
            "Epoch 4/5\n",
            "1175509/1175509 [==============================] - 66s 56us/sample - loss: 0.1159 - acc: 0.9548 - val_loss: 0.1183 - val_acc: 0.9541\n",
            "Epoch 5/5\n",
            "1175509/1175509 [==============================] - 66s 56us/sample - loss: 0.1121 - acc: 0.9558 - val_loss: 0.1184 - val_acc: 0.9547\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "paGYCXoqCMqf",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 182
        },
        "outputId": "94576847-2689-45ef-d32c-420157d8ebf6"
      },
      "source": [
        "#Determining the optimal threshold value to seggregate the classes\n",
        "pred4 = model4.predict(val_padded, batch_size=512,verbose=1)\n",
        "from sklearn import metrics\n",
        "for thresh in np.arange(0.1, 0.5, 0.05):\n",
        "    thresh = np.round(thresh, 2)\n",
        "    print(\"F1 score at threshold {0} is {1}\".format(thresh, metrics.f1_score(val_labels, (pred4>thresh).astype(int))))"
      ],
      "execution_count": 103,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "130613/130613 [==============================] - 3s 26us/sample\n",
            "F1 score at threshold 0.1 is 0.5517440184491208\n",
            "F1 score at threshold 0.15 is 0.5838901601830664\n",
            "F1 score at threshold 0.2 is 0.6016881827209534\n",
            "F1 score at threshold 0.25 is 0.610668366804803\n",
            "F1 score at threshold 0.3 is 0.6135282121006117\n",
            "F1 score at threshold 0.35 is 0.6142754145638067\n",
            "F1 score at threshold 0.4 is 0.6062571537581075\n",
            "F1 score at threshold 0.45 is 0.596748185971513\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YgvWgy9ICUx5",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 539
        },
        "outputId": "229e5a01-d0ec-4bd8-c83e-75f30f75657d"
      },
      "source": [
        "#From above, we can see that F1 score is highest at a threshold of 0.35\n",
        "pred_ft4_val_labels=(pred4>0.35).astype(int)\n",
        "\n",
        "from sklearn.metrics import classification_report\n",
        "print(classification_report(val_labels,pred_ft4_val_labels))\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "from sklearn.metrics import confusion_matrix\n",
        "print(confusion_matrix(val_labels, pred_ft4_val_labels))\n",
        "CM = confusion_matrix(val_labels, pred_ft4_val_labels)\n",
        "from mlxtend.plotting import plot_confusion_matrix\n",
        "fig, ax = plot_confusion_matrix(conf_mat=CM ,  figsize=(5, 5))\n",
        "plt.show()"
      ],
      "execution_count": 104,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.98      0.97      0.97    122534\n",
            "           1       0.60      0.63      0.61      8079\n",
            "\n",
            "    accuracy                           0.95    130613\n",
            "   macro avg       0.79      0.80      0.79    130613\n",
            "weighted avg       0.95      0.95      0.95    130613\n",
            "\n",
            "[[119081   3453]\n",
            " [  2967   5112]]\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAT0AAAFACAYAAADOJ6uCAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\npCcpK0ZPUlaMnqSsGD1JWTF6krLyPzznNHGFMjLjAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 360x360 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0II-hwtPolVu",
        "colab_type": "text"
      },
      "source": [
        "# **Final Model**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ISSEq_g2pTdf",
        "colab_type": "text"
      },
      "source": [
        "###**Though the results of the above models are similar, there is a good chance that they might capture different type of information from the data. So, the final model is a blend of the 4 previous models.**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mi5GpSlmhsuk",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 182
        },
        "outputId": "d54130cc-5fad-45d1-8bbc-2cd891f54c03"
      },
      "source": [
        "pred_final = 0.25*pred1 + 0.25*pred2 + 0.25*pred3 + 0.25*pred4\n",
        "for thresh in np.arange(0.1, 0.501, 0.05):\n",
        "    thresh = np.round(thresh, 2)\n",
        "    print(\"F1 score at threshold {0} is {1}\".format(thresh, metrics.f1_score(val_labels, (pred_final>thresh).astype(int))))"
      ],
      "execution_count": 116,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "F1 score at threshold 0.1 is 0.5465947764903828\n",
            "F1 score at threshold 0.15 is 0.5918959913326111\n",
            "F1 score at threshold 0.2 is 0.61951775204101\n",
            "F1 score at threshold 0.25 is 0.6376663254861823\n",
            "F1 score at threshold 0.3 is 0.6454759686523813\n",
            "F1 score at threshold 0.35 is 0.6486989929565167\n",
            "F1 score at threshold 0.4 is 0.6430995126164477\n",
            "F1 score at threshold 0.45 is 0.6308267226119452\n",
            "F1 score at threshold 0.5 is 0.611505780749408\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mrimbS99j34u",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 539
        },
        "outputId": "1795c656-c49f-4c66-9f5f-e3286d9eb153"
      },
      "source": [
        "pred_final_val_labels=(pred_final>0.33).astype(int)\n",
        "\n",
        "from sklearn.metrics import classification_report\n",
        "print(classification_report(val_labels,pred_final_val_labels))\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "from sklearn.metrics import confusion_matrix\n",
        "print(confusion_matrix(val_labels, pred_final_val_labels))\n",
        "CM = confusion_matrix(val_labels, pred_final_val_labels)\n",
        "from mlxtend.plotting import plot_confusion_matrix\n",
        "fig, ax = plot_confusion_matrix(conf_mat=CM ,  figsize=(5, 5))\n",
        "plt.show()"
      ],
      "execution_count": 118,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.98      0.97      0.97    122534\n",
            "           1       0.60      0.71      0.65      8079\n",
            "\n",
            "    accuracy                           0.95    130613\n",
            "   macro avg       0.79      0.84      0.81    130613\n",
            "weighted avg       0.96      0.95      0.95    130613\n",
            "\n",
            "[[118723   3811]\n",
            " [  2379   5700]]\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAT0AAAFACAYAAADOJ6uCAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 360x360 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "N_cyd77NvUqH",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 90
        },
        "outputId": "ad603070-22df-4423-92a1-2eb4743eab75"
      },
      "source": [
        "pred1_l = model.predict(test_padded, batch_size=512,verbose=1)\n",
        "pred2_l = model_gru.predict(test_padded, batch_size=512,verbose=1)\n",
        "pred3_l = model3_lstm.predict(test_padded, batch_size=512,verbose=1)\n",
        "pred4_l = model4.predict(test_padded, batch_size=512,verbose=1)\n",
        "pred_test_labels = 0.25*pred1_l + 0.25*pred2_l + 0.25*pred3_l + 0.25*pred4_l\n",
        "pred_test_labels = (pred_test_labels>0.35).astype(int)\n",
        "out_df = pd.DataFrame({'qid':df_test['qid'].values})\n",
        "out_df['prediction'] = pred_test_labels\n",
        "out_df.to_csv('submission.csv', index=False)"
      ],
      "execution_count": 122,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "375806/375806 [==============================] - 12s 32us/sample\n",
            "375806/375806 [==============================] - 11s 28us/sample\n",
            "375806/375806 [==============================] - 11s 30us/sample\n",
            "375806/375806 [==============================] - 10s 26us/sample\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-thIFr0qw3u9",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "out=pd.read_csv('submission.csv')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dOMSNoz-xHkC",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 72
        },
        "outputId": "6ae2a473-ee89-43ec-91a3-fe9e6e0a61ca"
      },
      "source": [
        "out['prediction'].value_counts()"
      ],
      "execution_count": 127,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0    349804\n",
              "1     26002\n",
              "Name: prediction, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 127
        }
      ]
    }
  ]
}