CNN+RNN-dropout.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "CNN",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/rachit-shah/Apparent-Age-Prediction-using-CNN/blob/master/CNN%2BRNN-dropout.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "metadata": {
        "id": "BToPi2JX6jzd",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Load Data and Glove Model From Drive"
      ]
    },
    {
      "metadata": {
        "id": "JvS3pf0t6zsk",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "e8cf924f-29ad-429e-f2e1-403f8dce6d8d"
      },
      "cell_type": "code",
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/gdrive')"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Drive already mounted at /content/gdrive; to attempt to forcibly remount, call drive.mount(\"/content/gdrive\", force_remount=True).\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "e9s7l5uI7QWJ",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 846
        },
        "outputId": "3471cebc-c029-47ef-adc7-4ebc9a1c327c"
      },
      "cell_type": "code",
      "source": [
        "import pandas as pd\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "import os\n",
        "%matplotlib inline\n",
        "import nltk\n",
        "nltk.download('popular')\n",
        "\n",
        "from keras.preprocessing.text import Tokenizer\n",
        "from keras.preprocessing.sequence import pad_sequences\n",
        "from keras.utils import to_categorical\n",
        "from keras.layers import Dense, Input, GlobalMaxPooling1D, Dropout, LSTM, GRU, Bidirectional, SpatialDropout1D\n",
        "from keras.layers import Conv1D, MaxPooling1D, Embedding, Flatten\n",
        "from keras.callbacks import ModelCheckpoint, EarlyStopping, TensorBoard\n",
        "from keras.models import Model\n",
        "from keras.initializers import Constant\n",
        "\n",
        "from wordcloud import WordCloud\n",
        "import matplotlib.pyplot as plt\n",
        "from sklearn.feature_extraction import text \n",
        "from wordcloud import WordCloud, STOPWORDS\n",
        "\n",
        "from sklearn.preprocessing import LabelEncoder\n",
        "import pickle\n",
        "import numpy as np\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "import nltk.corpus\n",
        "import string\n",
        "from nltk.tokenize import word_tokenize,sent_tokenize\n",
        "from nltk.stem import WordNetLemmatizer\n",
        "\n",
        "from keras.layers import Embedding\n",
        "\n",
        "import time"
      ],
      "execution_count": 82,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[nltk_data] Downloading collection 'popular'\n",
            "[nltk_data]    | \n",
            "[nltk_data]    | Downloading package cmudict to /root/nltk_data...\n",
            "[nltk_data]    |   Package cmudict is already up-to-date!\n",
            "[nltk_data]    | Downloading package gazetteers to /root/nltk_data...\n",
            "[nltk_data]    |   Package gazetteers is already up-to-date!\n",
            "[nltk_data]    | Downloading package genesis to /root/nltk_data...\n",
            "[nltk_data]    |   Package genesis is already up-to-date!\n",
            "[nltk_data]    | Downloading package gutenberg to /root/nltk_data...\n",
            "[nltk_data]    |   Package gutenberg is already up-to-date!\n",
            "[nltk_data]    | Downloading package inaugural to /root/nltk_data...\n",
            "[nltk_data]    |   Package inaugural is already up-to-date!\n",
            "[nltk_data]    | Downloading package movie_reviews to\n",
            "[nltk_data]    |     /root/nltk_data...\n",
            "[nltk_data]    |   Package movie_reviews is already up-to-date!\n",
            "[nltk_data]    | Downloading package names to /root/nltk_data...\n",
            "[nltk_data]    |   Package names is already up-to-date!\n",
            "[nltk_data]    | Downloading package shakespeare to /root/nltk_data...\n",
            "[nltk_data]    |   Package shakespeare is already up-to-date!\n",
            "[nltk_data]    | Downloading package stopwords to /root/nltk_data...\n",
            "[nltk_data]    |   Package stopwords is already up-to-date!\n",
            "[nltk_data]    | Downloading package treebank to /root/nltk_data...\n",
            "[nltk_data]    |   Package treebank is already up-to-date!\n",
            "[nltk_data]    | Downloading package twitter_samples to\n",
            "[nltk_data]    |     /root/nltk_data...\n",
            "[nltk_data]    |   Package twitter_samples is already up-to-date!\n",
            "[nltk_data]    | Downloading package omw to /root/nltk_data...\n",
            "[nltk_data]    |   Package omw is already up-to-date!\n",
            "[nltk_data]    | Downloading package wordnet to /root/nltk_data...\n",
            "[nltk_data]    |   Package wordnet is already up-to-date!\n",
            "[nltk_data]    | Downloading package wordnet_ic to /root/nltk_data...\n",
            "[nltk_data]    |   Package wordnet_ic is already up-to-date!\n",
            "[nltk_data]    | Downloading package words to /root/nltk_data...\n",
            "[nltk_data]    |   Package words is already up-to-date!\n",
            "[nltk_data]    | Downloading package maxent_ne_chunker to\n",
            "[nltk_data]    |     /root/nltk_data...\n",
            "[nltk_data]    |   Package maxent_ne_chunker is already up-to-date!\n",
            "[nltk_data]    | Downloading package punkt to /root/nltk_data...\n",
            "[nltk_data]    |   Package punkt is already up-to-date!\n",
            "[nltk_data]    | Downloading package snowball_data to\n",
            "[nltk_data]    |     /root/nltk_data...\n",
            "[nltk_data]    |   Package snowball_data is already up-to-date!\n",
            "[nltk_data]    | Downloading package averaged_perceptron_tagger to\n",
            "[nltk_data]    |     /root/nltk_data...\n",
            "[nltk_data]    |   Package averaged_perceptron_tagger is already up-\n",
            "[nltk_data]    |       to-date!\n",
            "[nltk_data]    | \n",
            "[nltk_data]  Done downloading collection popular\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "z8c7Ay-J7Qwh",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "!cp gdrive/'My Drive'/bbc-fulltext.zip .\n",
        "!cp gdrive/'My Drive'/glove.6B.zip .\n",
        "!unzip bbc-fulltext.zip > out.txt\n",
        "!rm bbc/README.TXT\n",
        "!unzip glove.6B.zip"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "Ew62XD1v790p",
        "colab_type": "code",
        "outputId": "af28a63c-aa70-476a-b86f-532525392f9a",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 198
        }
      },
      "cell_type": "code",
      "source": [
        "#Read Data\n",
        "categories = ['business','entertainment','politics','sport','tech']\n",
        "df = pd.DataFrame([],columns=['category','id','text'])\n",
        "for cat in categories:\n",
        "  for file in os.listdir(\"bbc/\"+cat):\n",
        "      if file.endswith(\".txt\"):\n",
        "          filepath = os.path.join(\"bbc/\"+cat, file)\n",
        "          text = open(filepath,'r', errors='ignore').read()\n",
        "          s = pd.Series([cat,int(filepath.split('/')[-1][:-4]),text],index=['category','id','text'])\n",
        "          df = df.append(s,ignore_index=True)\n",
        "\n",
        "df.head()"
      ],
      "execution_count": 3,
      "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>category</th>\n",
              "      <th>id</th>\n",
              "      <th>text</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>business</td>\n",
              "      <td>386</td>\n",
              "      <td>Krispy Kreme shares hit\\n\\nShares in Krispy Kr...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>business</td>\n",
              "      <td>190</td>\n",
              "      <td>Consumers drive French economy\\n\\nFrance's eco...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>business</td>\n",
              "      <td>439</td>\n",
              "      <td>German economy rebounds\\n\\nGermany's economy, ...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>business</td>\n",
              "      <td>454</td>\n",
              "      <td>LSE doubts boost bidders' shares\\n\\nShares in ...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>business</td>\n",
              "      <td>302</td>\n",
              "      <td>Cactus diet deal for Phytopharm\\n\\nA slimming ...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "   category   id                                               text\n",
              "0  business  386  Krispy Kreme shares hit\\n\\nShares in Krispy Kr...\n",
              "1  business  190  Consumers drive French economy\\n\\nFrance's eco...\n",
              "2  business  439  German economy rebounds\\n\\nGermany's economy, ...\n",
              "3  business  454  LSE doubts boost bidders' shares\\n\\nShares in ...\n",
              "4  business  302  Cactus diet deal for Phytopharm\\n\\nA slimming ..."
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 3
        }
      ]
    },
    {
      "metadata": {
        "id": "uZXIyNwn8tT3",
        "colab_type": "code",
        "outputId": "a459248a-5f9c-476f-fd07-1069f9ea6349",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 55
        }
      },
      "cell_type": "code",
      "source": [
        "'''\n",
        "#WordCloud\n",
        "cat = df['category'].unique()\n",
        "\n",
        "for i in range(0,len(cat)):\n",
        "    words = ' '.join(df.loc[df['category']==cat[i], 'text'])\n",
        "\n",
        "    wordcloud = WordCloud( \n",
        "                          stopwords=STOPWORDS,\n",
        "                          background_color='white',\n",
        "                          width=800,\n",
        "                          height=400\n",
        "                ).generate(words)\n",
        "    print(cat[i])\n",
        "    plt.figure(figsize=(10, 5))\n",
        "    plt.imshow(wordcloud)\n",
        "    plt.axis('off')\n",
        "    plt.show()\n",
        "\n",
        "words = ' '.join(df.loc[:, 'text'])\n",
        "\n",
        "wordcloud = WordCloud( \n",
        "                      stopwords=STOPWORDS,\n",
        "                      background_color='white',\n",
        "                      width=800,\n",
        "                      height=400\n",
        "            ).generate(words)\n",
        "print(\"ALL Categories:\")\n",
        "plt.figure(figsize=(10, 5))\n",
        "plt.imshow(wordcloud)\n",
        "plt.axis('off')\n",
        "plt.show()\n",
        "'''"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'\\n#WordCloud\\ncat = df[\\'category\\'].unique()\\n\\nfor i in range(0,len(cat)):\\n    words = \\' \\'.join(df.loc[df[\\'category\\']==cat[i], \\'text\\'])\\n\\n    wordcloud = WordCloud( \\n                          stopwords=STOPWORDS,\\n                          background_color=\\'white\\',\\n                          width=800,\\n                          height=400\\n                ).generate(words)\\n    print(cat[i])\\n    plt.figure(figsize=(10, 5))\\n    plt.imshow(wordcloud)\\n    plt.axis(\\'off\\')\\n    plt.show()\\n\\nwords = \\' \\'.join(df.loc[:, \\'text\\'])\\n\\nwordcloud = WordCloud( \\n                      stopwords=STOPWORDS,\\n                      background_color=\\'white\\',\\n                      width=800,\\n                      height=400\\n            ).generate(words)\\nprint(\"ALL Categories:\")\\nplt.figure(figsize=(10, 5))\\nplt.imshow(wordcloud)\\nplt.axis(\\'off\\')\\nplt.show()\\n'"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 4
        }
      ]
    },
    {
      "metadata": {
        "id": "JR2fsHlbr_Tj",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Label Encoder\n",
        "#### Use inverse_transform at the end after predicting"
      ]
    },
    {
      "metadata": {
        "id": "385fGoXNldB0",
        "colab_type": "code",
        "outputId": "cac8f4ef-2190-4e28-bf18-0b537db95d84",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 138
        }
      },
      "cell_type": "code",
      "source": [
        "y = df['category']\n",
        "X = df.drop(['category','id'],axis=1)\n",
        "X_train, X_test, y_train, y_test  = train_test_split(X,y,stratify=y, test_size=0.2, random_state=123)\n",
        "X_train = X_train.reset_index(drop=True)\n",
        "X_test = X_test.reset_index(drop=True)\n",
        "y_test = y_test.reset_index(drop=True)\n",
        "\n",
        "print('before: %s ...' %y_train[:5])\n",
        "\n",
        "le = LabelEncoder()\n",
        "le.fit(y_train)\n",
        "y_train = le.transform(y_train)\n",
        "\n",
        "print('after: %s ...' %y_train)"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "before: 1508       sport\n",
            "1908        tech\n",
            "2104        tech\n",
            "2019        tech\n",
            "1245    politics\n",
            "Name: category, dtype: object ...\n",
            "after: [3 4 4 ... 1 0 2] ...\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "n9jqvQtYDhR0",
        "colab_type": "code",
        "outputId": "6f0084e2-be0e-49ac-9876-cad5178f7cb8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 155
        }
      },
      "cell_type": "code",
      "source": [
        "print(np.bincount(y_train))\n",
        "print(y.value_counts())\n",
        "print(le.inverse_transform([i for i in range(5)]))"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[408 309 333 409 321]\n",
            "sport            511\n",
            "business         510\n",
            "politics         417\n",
            "tech             401\n",
            "entertainment    386\n",
            "Name: category, dtype: int64\n",
            "['business' 'entertainment' 'politics' 'sport' 'tech']\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "1e6g_BjOsZ3i",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Preprocess Data - (split by \\n, remove periods, remove slashes)"
      ]
    },
    {
      "metadata": {
        "id": "_3Iv6MC3s26i",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def preprocess(X):\n",
        "  return X.map(lambda x: x.lower().split(\"\\n\")).map(lambda x: [y.split(\". \") for y in x]).map(lambda x: [i.replace('\\'','') for sl in x for i in sl if i is not ''])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "6EgQMq_CtK8d",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "#Preprocess both train and test separately\n",
        "X_train['text'] = preprocess(X_train['text'])\n",
        "X_test['text'] = preprocess(X_test['text'])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "Wme2RkYu2UR0",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "#Stop Words and Lemmatization   \n",
        "lemmatizer = WordNetLemmatizer()\n",
        "stopw = nltk.corpus.stopwords.words('english')\n",
        "punct = string.punctuation\n",
        "punct = word_tokenize(punct)\n",
        "punct += ['.','``','...','\\'s','--','-','n\\'t','\\'']\n",
        "stopw += punct\n",
        "def token_stop(text):\n",
        "    global stopw\n",
        "    global lemmatizer\n",
        "    words = word_tokenize(text)\n",
        "    filtered = [lemmatizer.lemmatize(w) for w in words if not w in stopw]\n",
        "    return filtered\n",
        "  \n",
        "X_train['text'] = X_train['text'].map(lambda x: [token_stop(i) for i in x]).map(lambda x: [i for sl in x for i in sl])\n",
        "X_test['text'] = X_test['text'].map(lambda x: [token_stop(i) for i in x]).map(lambda x: [i for sl in x for i in sl])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "fHKBz-DG8HTu",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "texts = np.array(X_train['text'])\n",
        "test_text = np.array(X_test['text'])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "sLe602IIEqV6",
        "colab_type": "code",
        "outputId": "a4ec4fc8-4200-4688-c495-efe526390ff3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "cell_type": "code",
      "source": [
        "MAX_SEQUENCE_LENGTH = 1000\n",
        "MAX_NB_WORDS = 20000\n",
        "EMBEDDING_DIM = 100\n",
        "VALIDATION_SPLIT = 0.2\n",
        "embeddings_index = {}\n",
        "with open('glove.6B.100d.txt') as f:\n",
        "    for line in f:\n",
        "        word, coefs = line.split(maxsplit=1)\n",
        "        coefs = np.fromstring(coefs, 'f', sep=' ')\n",
        "        embeddings_index[word] = coefs\n",
        "\n",
        "print('Found %s word vectors.' % len(embeddings_index))"
      ],
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found 400000 word vectors.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "9thZOjdY_Sl6",
        "colab_type": "code",
        "outputId": "991f0b8f-0b2f-4c89-af7c-a433cbf5410a",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 124
        }
      },
      "cell_type": "code",
      "source": [
        "tokenizer = Tokenizer(nb_words=MAX_NB_WORDS)\n",
        "tokenizer.fit_on_texts(texts)\n",
        "sequences = tokenizer.texts_to_sequences(texts)\n",
        "\n",
        "word_index = tokenizer.word_index\n",
        "print('Found %s unique tokens.' % len(word_index))\n",
        "\n",
        "data = pad_sequences(sequences, maxlen=MAX_SEQUENCE_LENGTH)\n",
        "\n",
        "labels = to_categorical(np.asarray(y_train))\n",
        "print('Shape of data tensor:', data.shape)\n",
        "print('Shape of label tensor:', labels.shape)\n",
        "\n",
        "# split the data into a training set and a validation set\n",
        "np.random.seed(123)\n",
        "indices = np.arange(data.shape[0])\n",
        "np.random.shuffle(indices)\n",
        "data = data[indices]\n",
        "labels = labels[indices]\n",
        "nb_validation_samples = int(VALIDATION_SPLIT * data.shape[0])\n",
        "\n",
        "xtrain = data[:-nb_validation_samples]\n",
        "ytrain = labels[:-nb_validation_samples]\n",
        "xval = data[-nb_validation_samples:]\n",
        "yval = labels[-nb_validation_samples:]"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/keras_preprocessing/text.py:178: UserWarning: The `nb_words` argument in `Tokenizer` has been renamed `num_words`.\n",
            "  warnings.warn('The `nb_words` argument in `Tokenizer` '\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Found 28841 unique tokens.\n",
            "Shape of data tensor: (1780, 1000)\n",
            "Shape of label tensor: (1780, 5)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "VofHdiAqIRqc",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "#Create glove embedding matrix\n",
        "embedding_matrix = np.zeros((len(word_index) + 1, 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",
        "        # words not found in embedding index will be all-zeros.\n",
        "        embedding_matrix[i] = embedding_vector"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "o3yooq2DI-dT",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "#Keras Glove Embedding layer\n",
        "embedding_layer = Embedding(len(word_index) + 1,\n",
        "                            EMBEDDING_DIM,\n",
        "                            weights=[embedding_matrix],\n",
        "                            input_length=MAX_SEQUENCE_LENGTH,\n",
        "                            trainable=False)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "iuZyVOsKKD6T",
        "colab_type": "code",
        "outputId": "2b3edd6a-c1cc-4d21-99e0-ba8628e5c915",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "cell_type": "code",
      "source": [
        "#Map label name to its index\n",
        "labels_index = {}\n",
        "for i in range(5):\n",
        "  name = le.inverse_transform([i])[0]\n",
        "  labels_index[name] = i\n",
        "labels_index"
      ],
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{'business': 0, 'entertainment': 1, 'politics': 2, 'sport': 3, 'tech': 4}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 16
        }
      ]
    },
    {
      "metadata": {
        "id": "e18TBVrC9iwT",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Train Model"
      ]
    },
    {
      "metadata": {
        "id": "ijGeDWg6-XsZ",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "model_name = \"CNN\"\n",
        "checkpointer = ModelCheckpoint(model_name + \"_weights.{epoch:02d}-{val_loss:.2f}.hdf5\", monitor=\"val_loss\", verbose=1,\n",
        "                               save_best_only=True, mode='min')\n",
        "early_stopping = EarlyStopping(monitor='val_loss', min_delta=0, patience=10, verbose=1, mode='auto')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "53s9TZ_mI_XY",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "sequence_input = Input(shape=(MAX_SEQUENCE_LENGTH,), dtype='int32')\n",
        "embedded_sequences = embedding_layer(sequence_input)\n",
        "x = Conv1D(128, 5, activation='relu')(embedded_sequences)\n",
        "x = MaxPooling1D(5)(x)\n",
        "x = Dropout(0.2)(x)\n",
        "x = Conv1D(128, 5, activation='relu')(x)\n",
        "x = MaxPooling1D(5)(x)\n",
        "x = Dropout(0.2)(x)\n",
        "x = Conv1D(128, 5, activation='relu')(x)\n",
        "x = MaxPooling1D(35)(x)  # global max pooling\n",
        "x = Dropout(0.2)(x)\n",
        "x = Flatten()(x)\n",
        "x = Dense(128, activation='relu')(x)\n",
        "preds = Dense(len(labels_index), activation='softmax')(x)\n",
        "\n",
        "model = Model(sequence_input, preds)\n",
        "model.compile(loss='categorical_crossentropy',\n",
        "              optimizer='rmsprop',\n",
        "              metrics=['acc'])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "Fp3SQDq3-tfj",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1796
        },
        "outputId": "7a75cc63-4725-4d6c-f37c-dc6a91e8435e"
      },
      "cell_type": "code",
      "source": [
        "stmillis = int(round(time.time() * 1000))\n",
        "history = model.fit(xtrain, ytrain, validation_data=(xval, yval),\n",
        "          epochs=50, batch_size=128, callbacks=[checkpointer, early_stopping])\n",
        "endmillis = int(round(time.time() * 1000))\n",
        "print(\"Time taken: \", endmillis - stmillis)"
      ],
      "execution_count": 46,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 1424 samples, validate on 356 samples\n",
            "Epoch 1/50\n",
            "1424/1424 [==============================] - 16s 12ms/step - loss: 1.6922 - acc: 0.3385 - val_loss: 1.1786 - val_acc: 0.4270\n",
            "\n",
            "Epoch 00001: val_loss did not improve from 0.08955\n",
            "Epoch 2/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.8821 - acc: 0.6524 - val_loss: 0.7998 - val_acc: 0.6601\n",
            "\n",
            "Epoch 00002: val_loss did not improve from 0.08955\n",
            "Epoch 3/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.4334 - acc: 0.8455 - val_loss: 0.4807 - val_acc: 0.8371\n",
            "\n",
            "Epoch 00003: val_loss did not improve from 0.08955\n",
            "Epoch 4/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.2342 - acc: 0.9199 - val_loss: 1.0553 - val_acc: 0.6292\n",
            "\n",
            "Epoch 00004: val_loss did not improve from 0.08955\n",
            "Epoch 5/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.2996 - acc: 0.9024 - val_loss: 0.8322 - val_acc: 0.7697\n",
            "\n",
            "Epoch 00005: val_loss did not improve from 0.08955\n",
            "Epoch 6/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.2137 - acc: 0.9368 - val_loss: 0.1800 - val_acc: 0.9438\n",
            "\n",
            "Epoch 00006: val_loss did not improve from 0.08955\n",
            "Epoch 7/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.1500 - acc: 0.9466 - val_loss: 0.1296 - val_acc: 0.9607\n",
            "\n",
            "Epoch 00007: val_loss did not improve from 0.08955\n",
            "Epoch 8/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.1146 - acc: 0.9593 - val_loss: 0.0904 - val_acc: 0.9747\n",
            "\n",
            "Epoch 00008: val_loss did not improve from 0.08955\n",
            "Epoch 9/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.1481 - acc: 0.9515 - val_loss: 0.1155 - val_acc: 0.9635\n",
            "\n",
            "Epoch 00009: val_loss did not improve from 0.08955\n",
            "Epoch 10/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.0644 - acc: 0.9838 - val_loss: 0.0934 - val_acc: 0.9747\n",
            "\n",
            "Epoch 00010: val_loss did not improve from 0.08955\n",
            "Epoch 11/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.1290 - acc: 0.9572 - val_loss: 0.1228 - val_acc: 0.9663\n",
            "\n",
            "Epoch 00011: val_loss did not improve from 0.08955\n",
            "Epoch 12/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.0443 - acc: 0.9853 - val_loss: 0.0855 - val_acc: 0.9719\n",
            "\n",
            "Epoch 00012: val_loss improved from 0.08955 to 0.08546, saving model to CNN_weights.12-0.09.hdf5\n",
            "Epoch 13/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.0495 - acc: 0.9838 - val_loss: 0.0829 - val_acc: 0.9775\n",
            "\n",
            "Epoch 00013: val_loss improved from 0.08546 to 0.08288, saving model to CNN_weights.13-0.08.hdf5\n",
            "Epoch 14/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.0836 - acc: 0.9782 - val_loss: 0.1004 - val_acc: 0.9775\n",
            "\n",
            "Epoch 00014: val_loss did not improve from 0.08288\n",
            "Epoch 15/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.0183 - acc: 0.9965 - val_loss: 0.0825 - val_acc: 0.9803\n",
            "\n",
            "Epoch 00015: val_loss improved from 0.08288 to 0.08252, saving model to CNN_weights.15-0.08.hdf5\n",
            "Epoch 16/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.0271 - acc: 0.9902 - val_loss: 0.2016 - val_acc: 0.9494\n",
            "\n",
            "Epoch 00016: val_loss did not improve from 0.08252\n",
            "Epoch 17/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.0155 - acc: 0.9937 - val_loss: 0.1169 - val_acc: 0.9775\n",
            "\n",
            "Epoch 00017: val_loss did not improve from 0.08252\n",
            "Epoch 18/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.0471 - acc: 0.9874 - val_loss: 0.3696 - val_acc: 0.9270\n",
            "\n",
            "Epoch 00018: val_loss did not improve from 0.08252\n",
            "Epoch 19/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.0228 - acc: 0.9937 - val_loss: 0.0938 - val_acc: 0.9719\n",
            "\n",
            "Epoch 00019: val_loss did not improve from 0.08252\n",
            "Epoch 20/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.0050 - acc: 0.9986 - val_loss: 0.0981 - val_acc: 0.9747\n",
            "\n",
            "Epoch 00020: val_loss did not improve from 0.08252\n",
            "Epoch 21/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.0334 - acc: 0.9916 - val_loss: 0.1588 - val_acc: 0.9747\n",
            "\n",
            "Epoch 00021: val_loss did not improve from 0.08252\n",
            "Epoch 22/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.0074 - acc: 0.9986 - val_loss: 0.1112 - val_acc: 0.9719\n",
            "\n",
            "Epoch 00022: val_loss did not improve from 0.08252\n",
            "Epoch 23/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.0015 - acc: 1.0000 - val_loss: 0.1079 - val_acc: 0.9803\n",
            "\n",
            "Epoch 00023: val_loss did not improve from 0.08252\n",
            "Epoch 24/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.1670 - acc: 0.9754 - val_loss: 0.1766 - val_acc: 0.9635\n",
            "\n",
            "Epoch 00024: val_loss did not improve from 0.08252\n",
            "Epoch 25/50\n",
            "1424/1424 [==============================] - 16s 11ms/step - loss: 0.0042 - acc: 0.9986 - val_loss: 0.1196 - val_acc: 0.9803\n",
            "\n",
            "Epoch 00025: val_loss did not improve from 0.08252\n",
            "Epoch 00025: early stopping\n",
            "Time taken:  398997\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "Z73Z42jPBsd3",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 303
        },
        "outputId": "251c7973-ba9f-44c8-c231-8a24efb7311e"
      },
      "cell_type": "code",
      "source": [
        "fig=plt.figure()\n",
        "plt.plot(history.history['acc'],'r',linewidth=3.0)\n",
        "plt.plot(history.history['val_acc'],'b',linewidth=3.0)\n",
        "plt.legend(['Training Accuracy', 'Validation Accuracy'],fontsize=18)\n",
        "plt.xlabel('Epochs ',fontsize=16)\n",
        "plt.ylabel('Accuracy',fontsize=16)\n",
        "plt.title('Accuracy Curves : CNN',fontsize=16)\n",
        "fig.savefig('accuracy_cnn.png')\n",
        "plt.show()"
      ],
      "execution_count": 47,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYoAAAEdCAYAAAASHSDrAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nY4yJyBKFMcaYiP4fk1iHYMq2d7YAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "8nUIMMQi_KrR",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# Save the trained weights\n",
        "model.save(model_name + \".h5\")\n",
        "\n",
        "# Save model config as json\n",
        "model_json = model.to_json()\n",
        "with open(model_name + \".json\", \"w\") as json_file:\n",
        "    json_file.write(model_json)\n",
        "\n",
        "# Persist the model to your google drive [VERY IMPORTANT]\n",
        "!cp CNN.* gdrive/My\\ Drive/"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "7BAwr4x19nMo",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Predict on Test Data"
      ]
    },
    {
      "metadata": {
        "id": "vIBlzM5zFVhG",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def find_test_acc(prediction):\n",
        "  count = 0\n",
        "  t = 0\n",
        "  for pred in prediction:\n",
        "    p = pred.argmax()\n",
        "    name = le.inverse_transform([p])\n",
        "    if name == y_test.loc[count]:\n",
        "      t+=1\n",
        "    count+=1\n",
        "  print('Test Accuracy:',(t/count)*100,\"%\")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "1sM8QeKfJgRr",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def predict_test(model):\n",
        "  #Tokenize test data and apply model prediction\n",
        "  sequences = tokenizer.texts_to_sequences(test_text)\n",
        "  data = pad_sequences(sequences, maxlen=MAX_SEQUENCE_LENGTH)\n",
        "\n",
        "  print('Shape of data tensor:', data.shape)\n",
        "  print('Shape of label tensor:', labels.shape)\n",
        "  prediction = model.predict(data)\n",
        "  find_test_acc(prediction)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "u_z5eMImMj57",
        "colab_type": "code",
        "outputId": "62abb499-df73-4934-a930-1517dbc5456c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 69
        }
      },
      "cell_type": "code",
      "source": [
        "predict_test(model)"
      ],
      "execution_count": 62,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Shape of data tensor: (445, 1000)\n",
            "Shape of label tensor: (1780, 5)\n",
            "Test Accuracy: 98.87640449438202 %\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "p9D32NrbEPsU",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# RNN"
      ]
    },
    {
      "metadata": {
        "id": "NjPmDH75ERKA",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "sequence_input = Input(shape=(MAX_SEQUENCE_LENGTH,), dtype='int32')\n",
        "embedded_sequences = embedding_layer(sequence_input)\n",
        "x = SpatialDropout1D(0.2)(embedded_sequences)\n",
        "x = LSTM(100,dropout=0.2, recurrent_dropout=0.2)(x)\n",
        "preds = Dense(len(labels_index), activation='softmax')(x)\n",
        "lstm_model = Model(sequence_input, preds)\n",
        "lstm_model.compile(loss='categorical_crossentropy',\n",
        "              optimizer='rmsprop',\n",
        "              metrics=['acc'])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "dnLCsXllMNzi",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 311
        },
        "outputId": "a10b9a50-9450-4ac4-9676-9ccd0915b819"
      },
      "cell_type": "code",
      "source": [
        "lstm_model.summary()"
      ],
      "execution_count": 84,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_13 (InputLayer)        (None, 1000)              0         \n",
            "_________________________________________________________________\n",
            "embedding_1 (Embedding)      (None, 1000, 100)         2884200   \n",
            "_________________________________________________________________\n",
            "spatial_dropout1d_3 (Spatial (None, 1000, 100)         0         \n",
            "_________________________________________________________________\n",
            "lstm_5 (LSTM)                (None, 100)               80400     \n",
            "_________________________________________________________________\n",
            "dense_16 (Dense)             (None, 5)                 505       \n",
            "=================================================================\n",
            "Total params: 2,965,105\n",
            "Trainable params: 80,905\n",
            "Non-trainable params: 2,884,200\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "bbb9RTErE4wz",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "model_name = \"RNN\"\n",
        "checkpointer = ModelCheckpoint(model_name + \"_weights.{epoch:02d}-{val_loss:.2f}.hdf5\", monitor=\"val_loss\", verbose=1,\n",
        "                               save_best_only=True, mode='min')\n",
        "early_stopping = EarlyStopping(monitor='val_loss', min_delta=0, patience=10, verbose=1, mode='auto')\n",
        "tensorboard = TensorBoard(log_dir='/Graph', histogram_freq=0,write_graph=True, write_images=True)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "U9x73pzTEzaO",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 2695
        },
        "outputId": "526c119d-8667-4470-ac11-1a31877777a1"
      },
      "cell_type": "code",
      "source": [
        "stmillis = int(round(time.time() * 1000))\n",
        "history = lstm_model.fit(xtrain, ytrain, validation_data=(xval, yval),\n",
        "          epochs=50, batch_size=128, callbacks=[checkpointer, early_stopping,tensorboard])\n",
        "endmillis = int(round(time.time() * 1000))\n",
        "print(\"Time taken: \", endmillis - stmillis)"
      ],
      "execution_count": 87,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 1424 samples, validate on 356 samples\n",
            "Epoch 1/50\n",
            "1424/1424 [==============================] - 44s 31ms/step - loss: 1.3346 - acc: 0.5260 - val_loss: 0.7390 - val_acc: 0.7669\n",
            "\n",
            "Epoch 00001: val_loss improved from inf to 0.73901, saving model to RNN_weights.01-0.74.hdf5\n",
            "Epoch 2/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.8155 - acc: 0.7177 - val_loss: 0.4868 - val_acc: 0.8483\n",
            "\n",
            "Epoch 00002: val_loss improved from 0.73901 to 0.48683, saving model to RNN_weights.02-0.49.hdf5\n",
            "Epoch 3/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.6508 - acc: 0.7900 - val_loss: 0.3774 - val_acc: 0.8848\n",
            "\n",
            "Epoch 00003: val_loss improved from 0.48683 to 0.37741, saving model to RNN_weights.03-0.38.hdf5\n",
            "Epoch 4/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.5510 - acc: 0.8272 - val_loss: 0.3632 - val_acc: 0.8792\n",
            "\n",
            "Epoch 00004: val_loss improved from 0.37741 to 0.36321, saving model to RNN_weights.04-0.36.hdf5\n",
            "Epoch 5/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.5731 - acc: 0.8195 - val_loss: 0.3134 - val_acc: 0.9101\n",
            "\n",
            "Epoch 00005: val_loss improved from 0.36321 to 0.31342, saving model to RNN_weights.05-0.31.hdf5\n",
            "Epoch 6/50\n",
            "1424/1424 [==============================] - 43s 31ms/step - loss: 0.4640 - acc: 0.8469 - val_loss: 0.3016 - val_acc: 0.9213\n",
            "\n",
            "Epoch 00006: val_loss improved from 0.31342 to 0.30159, saving model to RNN_weights.06-0.30.hdf5\n",
            "Epoch 7/50\n",
            "1424/1424 [==============================] - 44s 31ms/step - loss: 0.4225 - acc: 0.8743 - val_loss: 0.2184 - val_acc: 0.9382\n",
            "\n",
            "Epoch 00007: val_loss improved from 0.30159 to 0.21837, saving model to RNN_weights.07-0.22.hdf5\n",
            "Epoch 8/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.4024 - acc: 0.8701 - val_loss: 0.2385 - val_acc: 0.9354\n",
            "\n",
            "Epoch 00008: val_loss did not improve from 0.21837\n",
            "Epoch 9/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.4308 - acc: 0.8610 - val_loss: 0.2084 - val_acc: 0.9522\n",
            "\n",
            "Epoch 00009: val_loss improved from 0.21837 to 0.20838, saving model to RNN_weights.09-0.21.hdf5\n",
            "Epoch 10/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.3802 - acc: 0.8848 - val_loss: 0.2370 - val_acc: 0.9298\n",
            "\n",
            "Epoch 00010: val_loss did not improve from 0.20838\n",
            "Epoch 11/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.3364 - acc: 0.8912 - val_loss: 0.1741 - val_acc: 0.9522\n",
            "\n",
            "Epoch 00011: val_loss improved from 0.20838 to 0.17412, saving model to RNN_weights.11-0.17.hdf5\n",
            "Epoch 12/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.3343 - acc: 0.8968 - val_loss: 0.2338 - val_acc: 0.9326\n",
            "\n",
            "Epoch 00012: val_loss did not improve from 0.17412\n",
            "Epoch 13/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.3709 - acc: 0.8926 - val_loss: 0.1793 - val_acc: 0.9551\n",
            "\n",
            "Epoch 00013: val_loss did not improve from 0.17412\n",
            "Epoch 14/50\n",
            "1424/1424 [==============================] - 44s 31ms/step - loss: 0.3349 - acc: 0.8926 - val_loss: 0.2348 - val_acc: 0.9382\n",
            "\n",
            "Epoch 00014: val_loss did not improve from 0.17412\n",
            "Epoch 15/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.3436 - acc: 0.8968 - val_loss: 0.1615 - val_acc: 0.9579\n",
            "\n",
            "Epoch 00015: val_loss improved from 0.17412 to 0.16153, saving model to RNN_weights.15-0.16.hdf5\n",
            "Epoch 16/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.3090 - acc: 0.9024 - val_loss: 0.1912 - val_acc: 0.9466\n",
            "\n",
            "Epoch 00016: val_loss did not improve from 0.16153\n",
            "Epoch 17/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.3182 - acc: 0.9038 - val_loss: 0.1779 - val_acc: 0.9494\n",
            "\n",
            "Epoch 00017: val_loss did not improve from 0.16153\n",
            "Epoch 18/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.3278 - acc: 0.8940 - val_loss: 0.1508 - val_acc: 0.9607\n",
            "\n",
            "Epoch 00018: val_loss improved from 0.16153 to 0.15083, saving model to RNN_weights.18-0.15.hdf5\n",
            "Epoch 19/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.2787 - acc: 0.9164 - val_loss: 0.2534 - val_acc: 0.9242\n",
            "\n",
            "Epoch 00019: val_loss did not improve from 0.15083\n",
            "Epoch 20/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.2754 - acc: 0.9192 - val_loss: 0.1869 - val_acc: 0.9466\n",
            "\n",
            "Epoch 00020: val_loss did not improve from 0.15083\n",
            "Epoch 21/50\n",
            "1424/1424 [==============================] - 44s 31ms/step - loss: 0.2738 - acc: 0.9199 - val_loss: 0.1500 - val_acc: 0.9635\n",
            "\n",
            "Epoch 00021: val_loss improved from 0.15083 to 0.14996, saving model to RNN_weights.21-0.15.hdf5\n",
            "Epoch 22/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.3188 - acc: 0.8989 - val_loss: 0.1878 - val_acc: 0.9410\n",
            "\n",
            "Epoch 00022: val_loss did not improve from 0.14996\n",
            "Epoch 23/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.2803 - acc: 0.9129 - val_loss: 0.1376 - val_acc: 0.9579\n",
            "\n",
            "Epoch 00023: val_loss improved from 0.14996 to 0.13764, saving model to RNN_weights.23-0.14.hdf5\n",
            "Epoch 24/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.2732 - acc: 0.9171 - val_loss: 0.2553 - val_acc: 0.9185\n",
            "\n",
            "Epoch 00024: val_loss did not improve from 0.13764\n",
            "Epoch 25/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.2622 - acc: 0.9235 - val_loss: 0.2014 - val_acc: 0.9438\n",
            "\n",
            "Epoch 00025: val_loss did not improve from 0.13764\n",
            "Epoch 26/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.2597 - acc: 0.9199 - val_loss: 0.1857 - val_acc: 0.9466\n",
            "\n",
            "Epoch 00026: val_loss did not improve from 0.13764\n",
            "Epoch 27/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.2497 - acc: 0.9242 - val_loss: 0.1556 - val_acc: 0.9579\n",
            "\n",
            "Epoch 00027: val_loss did not improve from 0.13764\n",
            "Epoch 28/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.2491 - acc: 0.9235 - val_loss: 0.1314 - val_acc: 0.9607\n",
            "\n",
            "Epoch 00028: val_loss improved from 0.13764 to 0.13135, saving model to RNN_weights.28-0.13.hdf5\n",
            "Epoch 29/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.2164 - acc: 0.9375 - val_loss: 0.1640 - val_acc: 0.9522\n",
            "\n",
            "Epoch 00029: val_loss did not improve from 0.13135\n",
            "Epoch 30/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.2113 - acc: 0.9368 - val_loss: 0.1390 - val_acc: 0.9579\n",
            "\n",
            "Epoch 00030: val_loss did not improve from 0.13135\n",
            "Epoch 31/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.2218 - acc: 0.9291 - val_loss: 0.1344 - val_acc: 0.9635\n",
            "\n",
            "Epoch 00031: val_loss did not improve from 0.13135\n",
            "Epoch 32/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.2345 - acc: 0.9242 - val_loss: 0.1741 - val_acc: 0.9522\n",
            "\n",
            "Epoch 00032: val_loss did not improve from 0.13135\n",
            "Epoch 33/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.2062 - acc: 0.9347 - val_loss: 0.1606 - val_acc: 0.9494\n",
            "\n",
            "Epoch 00033: val_loss did not improve from 0.13135\n",
            "Epoch 34/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.2008 - acc: 0.9403 - val_loss: 0.1372 - val_acc: 0.9607\n",
            "\n",
            "Epoch 00034: val_loss did not improve from 0.13135\n",
            "Epoch 35/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.2343 - acc: 0.9312 - val_loss: 0.1318 - val_acc: 0.9663\n",
            "\n",
            "Epoch 00035: val_loss did not improve from 0.13135\n",
            "Epoch 36/50\n",
            "1424/1424 [==============================] - 44s 31ms/step - loss: 0.2018 - acc: 0.9333 - val_loss: 0.1773 - val_acc: 0.9410\n",
            "\n",
            "Epoch 00036: val_loss did not improve from 0.13135\n",
            "Epoch 37/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.1996 - acc: 0.9326 - val_loss: 0.1550 - val_acc: 0.9579\n",
            "\n",
            "Epoch 00037: val_loss did not improve from 0.13135\n",
            "Epoch 38/50\n",
            "1424/1424 [==============================] - 43s 30ms/step - loss: 0.1749 - acc: 0.9417 - val_loss: 0.1391 - val_acc: 0.9635\n",
            "\n",
            "Epoch 00038: val_loss did not improve from 0.13135\n",
            "Epoch 00038: early stopping\n",
            "Time taken:  1650108\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "vl8sxWfCEvhn",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 303
        },
        "outputId": "e99061ab-b9da-481f-ed01-f550fed7d944"
      },
      "cell_type": "code",
      "source": [
        "fig=plt.figure()\n",
        "plt.plot(history.history['acc'],'r',linewidth=3.0)\n",
        "plt.plot(history.history['val_acc'],'b',linewidth=3.0)\n",
        "plt.legend(['Training Accuracy', 'Validation Accuracy'],fontsize=18)\n",
        "plt.xlabel('Epochs ',fontsize=16)\n",
        "plt.ylabel('Accuracy',fontsize=16)\n",
        "plt.title('Accuracy Curves : CNN',fontsize=16)\n",
        "fig.savefig('accuracy_cnn.png')\n",
        "plt.show()"
      ],
      "execution_count": 88,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYoAAAEdCAYAAAASHSDrAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "VbE3Cy1LFJhy",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# Save the trained weights\n",
        "model.save(model_name + \".h5\")\n",
        "\n",
        "# Save model config as json\n",
        "model_json = model.to_json()\n",
        "with open(model_name + \".json\", \"w\") as json_file:\n",
        "    json_file.write(model_json)\n",
        "\n",
        "# Persist the model to your google drive [VERY IMPORTANT]\n",
        "!cp RNN.* gdrive/My\\ Drive/"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "6dqZeS3tGII0",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 69
        },
        "outputId": "1add5b06-e0d6-4373-dcb1-a6df940498e3"
      },
      "cell_type": "code",
      "source": [
        "predict_test(lstm_model)"
      ],
      "execution_count": 90,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Shape of data tensor: (445, 1000)\n",
            "Shape of label tensor: (1780, 5)\n",
            "Test Accuracy: 97.52808988764045 %\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "Bz0qLhb1Gi2s",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}