Category_Prediction_CV_Lemma.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.7.1"
    },
    "colab": {
      "name": "Category_Prediction_CV_Lemma.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "machine_shape": "hm",
      "include_colab_link": true
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/tpmarsha/ML2AmazonKaggle/blob/master/Category_Prediction_CV_Lemma.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_-x9J3S1khRx",
        "colab_type": "code",
        "outputId": "2694755d-faaa-448b-ac66-03b64ddf83ae",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 50
        }
      },
      "source": [
        "import pandas as pd\n",
        "import os\n",
        "from collections import defaultdict\n",
        "import time\n",
        "import numpy as np\n",
        "from numpy import random\n",
        "import nltk\n",
        "nltk.download('stopwords')\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.feature_extraction.text import CountVectorizer, TfidfVectorizer\n",
        "from sklearn.metrics import accuracy_score, classification_report\n",
        "from sklearn.pipeline import Pipeline\n",
        "from sklearn.feature_extraction.text import TfidfTransformer\n",
        "from sklearn.linear_model import LogisticRegression\n",
        "from sklearn.model_selection import cross_validate\n",
        "import matplotlib.pyplot as plt\n",
        "from nltk.corpus import stopwords\n",
        "import re\n",
        "\n",
        "\n"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[nltk_data] Downloading package stopwords to /root/nltk_data...\n",
            "[nltk_data]   Package stopwords is already up-to-date!\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "FX_SmV8AkhR1",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def read_json(f):\n",
        "  for l in open(f,\"r\"):\n",
        "    yield eval(l)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3NF2Gf8akhR3",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# f = \"C:/Users/Hewitt/Documents/GitHub/ML2_FinalProject/train.json\"\n",
        "# labels = ['Men','Women','Baby', 'Boys','Girls']\n",
        "# for l in read_json(f):\n",
        "#     start_time = time.time()\n",
        "#     reviewHash, reviewText, category = l['reviewHash'],l['reviewText'],l['categories']\n",
        "#     flat_category = [item for sublist in category for item in sublist]\n",
        "#     label = [label for label in labels if\n",
        "#              any(label in str for str in flat_category)] \n",
        "# #     print(label[0])\n",
        "#     df = df.append({'ReviewHash': reviewHash,\n",
        "#                     'ReviewText': reviewText,\n",
        "#                    'CategoryLabel': label[0]}, ignore_index=True)\n",
        "# #     print(\"--- %s seconds ---\" % (time.time() - start_time))\n",
        "# df.to_csv(\"train.csv\")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "G2ZDFVfVk_sV",
        "colab_type": "code",
        "outputId": "4a8c0a16-2ece-4bf4-fb9d-96e48f1c9d09",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 121
        }
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "execution_count": 5,
      "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&response_type=code&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\n",
            "\n",
            "Enter your authorization code:\n",
            "··········\n",
            "Mounted at /content/drive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ruaOnkJskhR5",
        "colab_type": "code",
        "outputId": "c30fa209-c1d3-4da2-bfed-4d8eeee291ca",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 195
        }
      },
      "source": [
        "root_path = '/content/drive/My Drive/Team 3 Kaggle Competition ML2/'\n",
        "df= pd.read_csv(root_path + 'train_CategoryPrediction.csv',\n",
        "               index_col=0)\n",
        "df.head(5)\n"
      ],
      "execution_count": 6,
      "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>ReviewHash</th>\n",
              "      <th>ReviewText</th>\n",
              "      <th>CategoryLabel</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>R798569390</td>\n",
              "      <td>The model in this picture has them rolled up a...</td>\n",
              "      <td>Women</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>R436443063</td>\n",
              "      <td>I love the look of this bra, it is what I want...</td>\n",
              "      <td>Women</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>R103439446</td>\n",
              "      <td>I am not comfortable with wearing my wedding b...</td>\n",
              "      <td>Women</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>R486351639</td>\n",
              "      <td>Like the look of this top and really looks cut...</td>\n",
              "      <td>Women</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>R508664275</td>\n",
              "      <td>I'm quite small and the XS fits me like a regu...</td>\n",
              "      <td>Women</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "   ReviewHash                                         ReviewText CategoryLabel\n",
              "0  R798569390  The model in this picture has them rolled up a...         Women\n",
              "1  R436443063  I love the look of this bra, it is what I want...         Women\n",
              "2  R103439446  I am not comfortable with wearing my wedding b...         Women\n",
              "3  R486351639  Like the look of this top and really looks cut...         Women\n",
              "4  R508664275  I'm quite small and the XS fits me like a regu...         Women"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ew6OP3bQkhR-",
        "colab_type": "code",
        "outputId": "154ef9e1-1599-4d34-ed58-8315fece5051",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 118
        }
      },
      "source": [
        "display(df['CategoryLabel'].value_counts())"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Women    141303\n",
              "Men       51530\n",
              "Baby       3324\n",
              "Girls      2191\n",
              "Boys       1652\n",
              "Name: CategoryLabel, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JpKszWGwkhSA",
        "colab_type": "code",
        "outputId": "52cfa376-e6e4-4d5c-a12e-cebc79e2cb0a",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 185
        }
      },
      "source": [
        "Replace_by_space_re = re.compile('[/(){}\\[\\]\\|@,;]')\n",
        "Bad_symbols_re = re.compile('[^0-9a-z #+_]')\n",
        "Stopwords = set(stopwords.words('english'))\n",
        "\n",
        "def clean_text(text):\n",
        "    text=str(text)\n",
        "    text=text.lower()\n",
        "    text=Replace_by_space_re.sub(' ', text)\n",
        "    text=Bad_symbols_re.sub('', text)\n",
        "    text= ' '.join(word for word in text.split() if word not in Stopwords)\n",
        "    return text\n",
        "\n",
        "df['ReviewText'] = df['ReviewText'].apply(clean_text)\n",
        "print(df.head(5))\n",
        "### Word Size\n",
        "print(df['ReviewText'].apply(lambda x: len(x.split(' '))).sum())\n",
        "### Max Review Length\n",
        "print(df['ReviewText'].apply(lambda x: len(x.split(' '))).max())\n",
        "\n",
        "df['tokens_length'] = df['ReviewText'].apply(lambda x: len(x.split(' ')))\n",
        "df[df.tokens_length==2046].ReviewText\n"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "   ReviewHash                                         ReviewText CategoryLabel\n",
            "0  R798569390  model picture rolled top actually high waisted...         Women\n",
            "1  R436443063  love look bra wanted however cup size band siz...         Women\n",
            "2  R103439446  comfortable wearing wedding band little diamon...         Women\n",
            "3  R486351639  like look top really looks cute buton closer l...         Women\n",
            "4  R508664275  im quite small xs fits like regular tshirt imp...         Women\n",
            "5889335\n",
            "2046\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "164763    kind icebreaker oasis shirt last week decided ...\n",
              "Name: ReviewText, dtype: object"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JeJDCfmmkhSC",
        "colab_type": "code",
        "outputId": "2e08ed21-ddf7-402a-f1aa-e14a7b70311e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 458
        }
      },
      "source": [
        "### Generate words count & histograms\n",
        "from yellowbrick.text import FreqDistVisualizer\n",
        "v=CountVectorizer()\n",
        "d=v.fit_transform(df['ReviewText'])\n",
        "f=v.get_feature_names()\n",
        "\n",
        "visualizer = FreqDistVisualizer(features=f, orient='v', n=2500)\n",
        "visualizer.fit(d)\n",
        "visualizer.draw()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/sklearn/utils/deprecation.py:144: FutureWarning: The sklearn.metrics.classification module is  deprecated in version 0.22 and will be removed in version 0.24. The corresponding classes / functions should instead be imported from sklearn.metrics. Anything that cannot be imported from sklearn.metrics is now part of the private API.\n",
            "  warnings.warn(message, FutureWarning)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 576x396 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LlZF6sD5khSF",
        "colab_type": "code",
        "outputId": "efc922ea-0261-49de-a6d5-ead32f158ddb",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 134
        }
      },
      "source": [
        "from nltk import FreqDist\n",
        "df['tokenized'] = df['ReviewText'].apply(lambda x: x.split(' '))\n",
        "# df.head(5)\n",
        "all_words = [word for item in list(df['tokenized']) for word in item]\n",
        "fdist = FreqDist(all_words)\n",
        "print(len(fdist)) # number of unique words\n",
        "# choose k and visually inspect the bottom 10 words of the top k\n",
        "k = 10000\n",
        "top_k_words = fdist.most_common(k)\n",
        "print(top_k_words[-10:])\n",
        "\n",
        "k = 5000\n",
        "top_k_words = fdist.most_common(k)\n",
        "print(top_k_words[-10:])\n",
        "\n",
        "k = 2000\n",
        "top_k_words = fdist.most_common(k)\n",
        "print(top_k_words[-10:])\n",
        "\n",
        "k = 1000\n",
        "top_k_words = fdist.most_common(k)\n",
        "print(top_k_words[-10:])\n",
        "\n",
        "k = 500\n",
        "top_k_words = fdist.most_common(k)\n",
        "print(top_k_words[-10:])\n",
        "\n",
        "k = 100\n",
        "top_k_words = fdist.most_common(k)\n",
        "print(top_k_words[-10:])\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "123375\n",
            "[('toesi', 19), ('spd', 19), ('44#34', 19), ('gifti', 19), ('braces', 19), ('poolside', 19), ('understandable', 19), ('1314', 19), ('singing', 19), ('approve', 19)]\n",
            "[('tracking', 67), ('handled', 67), ('eternity', 67), ('dishes', 67), ('pilates', 67), ('reads', 67), ('http', 67), ('score', 67), ('kudos', 67), ('neoprene', 67)]\n",
            "[('satin', 320), ('smart', 320), ('manufacturers', 320), ('def', 319), ('mark', 319), ('dr', 318), ('beginning', 318), ('amazoncom', 318), ('charge', 318), ('rolled', 317)]\n",
            "[('saying', 888), ('colored', 887), ('nicer', 885), ('figured', 884), ('skechers', 883), ('16', 881), ('rough', 877), ('flexible', 876), ('collar', 872), ('affordable', 872)]\n",
            "[('hoping', 2154), ('making', 2152), ('women', 2148), ('sock', 2146), ('dresses', 2145), ('womens', 2142), ('okay', 2136), ('zipper', 2119), ('pain', 2118), ('month', 2099)]\n",
            "[('style', 10241), ('lot', 10235), ('fabric', 10032), ('two', 9959), ('put', 9898), ('keep', 9890), ('cant', 9868), ('worn', 9802), ('since', 9772), ('true', 9711)]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Rn9b9ibEkhSJ",
        "colab_type": "code",
        "outputId": "4b39dc1b-0eab-45b0-87af-5d01abbaa9aa",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 84
        }
      },
      "source": [
        "# define a function only to keep words in the top k words\n",
        "k=2500\n",
        "top_k_words,_ = zip(*fdist.most_common(k))\n",
        "top_k_words = set(top_k_words)\n",
        "def keep_top_k_words(text):\n",
        "    return [word for word in text if word in top_k_words]\n",
        "\n",
        "df['tokenized'] = df['tokenized'].apply(keep_top_k_words)\n",
        "\n",
        "# document length\n",
        "df['doc_len'] = df['tokenized'].apply(lambda x: len(x))\n",
        "doc_lengths = list(df['doc_len'])\n",
        "\n",
        "print(\"length of list:\",len(doc_lengths),\n",
        "      \"\\naverage document length\", np.average(doc_lengths),\n",
        "      \"\\nminimum document length\", min(doc_lengths),\n",
        "      \"\\nmaximum document length\", max(doc_lengths))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "length of list: 200000 \n",
            "average document length 25.805625 \n",
            "minimum document length 0 \n",
            "maximum document length 1554\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kHM8sjm1khSL",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 202
        },
        "outputId": "4ca871de-e33b-4648-c33a-9b8e27072b6f"
      },
      "source": [
        "X =df.ReviewText\n",
        "print(X.head(10))\n",
        "y= df.CategoryLabel\n",
        "# X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state = 42, stratify=y)"
      ],
      "execution_count": 36,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "0    model picture rolled top actually high waisted...\n",
            "1    love look bra wanted however cup size band siz...\n",
            "2    comfortable wearing wedding band little diamon...\n",
            "3    like look top really looks cute buton closer l...\n",
            "4    im quite small xs fits like regular tshirt imp...\n",
            "5    bra help extra large girls 44ddd really well e...\n",
            "6    purchased shoes gift grand daughter loves fran...\n",
            "7    really enjoyed heat holders thermal socks tend...\n",
            "8    elastic sturdy taut tight shoe comfortable hig...\n",
            "9    due reviews back bought dress ordered size lik...\n",
            "Name: ReviewText, dtype: object\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "FfQbXlbnq9Ti",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 151
        },
        "outputId": "a68b4e37-198e-463c-a1bd-6a1b0e6926b1"
      },
      "source": [
        "from nltk.stem import WordNetLemmatizer\n",
        "nltk.download('wordnet')\n",
        "wnl = WordNetLemmatizer()\n",
        "def stemming(text):  \n",
        "  t = ' '.join(wnl.lemmatize(word) for word in text.split())\n",
        "  return t\n",
        "stemmed_X = X.apply(stemming)\n",
        "stemmed_X.head(5)"
      ],
      "execution_count": 35,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[nltk_data] Downloading package wordnet to /root/nltk_data...\n",
            "[nltk_data]   Unzipping corpora/wordnet.zip.\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0    model picture rolled top actually high waisted...\n",
              "1    love look bra wanted however cup size band siz...\n",
              "2    comfortable wearing wedding band little diamon...\n",
              "3    like look top really look cute buton closer lo...\n",
              "4    im quite small x fit like regular tshirt impor...\n",
              "Name: ReviewText, dtype: object"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 35
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pLOrIzkDkhSN",
        "colab_type": "code",
        "outputId": "ceb33457-c712-47f5-f4ce-b9b8f15fa356",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 104
        }
      },
      "source": [
        "tfidf = Pipeline([('vect', CountVectorizer()),\n",
        "                 ('tfidf', TfidfTransformer()),\n",
        "                 ('clf', LogisticRegression(n_jobs=10, C=1e5, multi_class='auto', verbose=1)),])\n",
        "start_time = time.time()\n",
        "scores = cross_validate(tfidf,stemmed_X,y,cv=5,n_jobs=10,verbose=1,return_train_score=True,scoring='accuracy')\n",
        "print(scores)\n",
        "print(\"--- %s seconds ---\" % (time.time() - start_time))\n",
        "\n"
      ],
      "execution_count": 38,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[Parallel(n_jobs=10)]: Using backend LokyBackend with 10 concurrent workers.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "{'fit_time': array([68.0115335 , 66.16965246, 66.56484389, 66.66375613, 66.98855877]), 'score_time': array([2.69778728, 2.81516647, 2.62171435, 2.60170364, 2.96751618]), 'test_score': array([0.83815 , 0.836825, 0.833975, 0.835425, 0.832175]), 'train_score': array([0.881075  , 0.8777125 , 0.878025  , 0.8769875 , 0.87585625])}\n",
            "--- 84.28455781936646 seconds ---\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[Parallel(n_jobs=10)]: Done   5 out of   5 | elapsed:  1.4min finished\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OwnfAbQRkhSQ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from tensorflow.keras.preprocessing.text import Tokenizer\n",
        "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
        "from tensorflow.keras.utils import to_categorical\n",
        "n_most_common_words = 2500 # based on the inspection of the words counts above\n",
        "max_len = 1000 # based on max document length above after filtering for top 2500 words\n",
        "tokenizer = Tokenizer(num_words=n_most_common_words)\n",
        "tokenizer.fit_on_texts(X)\n",
        "sequences = tokenizer.texts_to_sequences(X)\n",
        "X_LSTM = pad_sequences(sequences, maxlen=max_len)\n",
        "y_LSTM = to_categorical(df.CategoryLabel.factorize()[0], num_classes=5)\n",
        "X_train_lstm, X_test_lstm, y_train_lstm, y_test_lstm = train_test_split(X_LSTM, y_LSTM, test_size=0.2, random_state = 42, stratify=y)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mJ63dFG8khST",
        "colab_type": "code",
        "outputId": "d5a7480f-68ec-437c-fed6-be4f1e46363c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 319
        }
      },
      "source": [
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "from tensorflow.keras import Sequential\n",
        "from tensorflow.keras.layers import Embedding, Dense, Bidirectional, LSTM\n",
        "\n",
        "embedding_size = 100\n",
        "model = Sequential()\n",
        "model.add(Embedding(n_most_common_words, embedding_size, input_length=X_LSTM.shape[1]))\n",
        "model.add(Bidirectional(LSTM(128, dropout=0.7, recurrent_dropout=0.7)))\n",
        "model.add(Dense(y_LSTM.shape[1], activation='softmax'))\n",
        "model.compile(optimizer='adam', \n",
        "              loss='categorical_crossentropy',\n",
        "              metrics=['accuracy'])\n",
        "print(model.summary())"
      ],
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:Layer lstm will not use cuDNN kernel since it doesn't meet the cuDNN kernel criteria. It will use generic GPU kernel as fallback when running on GPU\n",
            "WARNING:tensorflow:Layer lstm will not use cuDNN kernel since it doesn't meet the cuDNN kernel criteria. It will use generic GPU kernel as fallback when running on GPU\n",
            "WARNING:tensorflow:Layer lstm will not use cuDNN kernel since it doesn't meet the cuDNN kernel criteria. It will use generic GPU kernel as fallback when running on GPU\n",
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "embedding (Embedding)        (None, 1000, 100)         250000    \n",
            "_________________________________________________________________\n",
            "bidirectional (Bidirectional (None, 256)               234496    \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, 5)                 1285      \n",
            "=================================================================\n",
            "Total params: 485,781\n",
            "Trainable params: 485,781\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "None\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vD5H-rzTkhSV",
        "colab_type": "code",
        "outputId": "fbba4ba7-c270-4c27-9328-15200f00cab7",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 67
        }
      },
      "source": [
        "start_time = time.time()\n",
        "model.fit(X_train_lstm, y_train_lstm, batch_size=128, validation_split=0.2, verbose=1, \n",
        "          shuffle=True, workers=10, use_multiprocessing=True)\n",
        "model.evaluate(X_test_lstm,y_test_lstm, verbose=1, workers=10, use_multiprocessing=True)\n",
        "\n",
        "print(\"--- %s seconds ---\" % (time.time() - start_time))\n"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "1000/1000 [==============================] - 3445s 3s/step - loss: 0.5545 - accuracy: 0.7899 - val_loss: 0.4555 - val_accuracy: 0.8270\n",
            "1250/1250 [==============================] - 652s 522ms/step - loss: 0.4607 - accuracy: 0.8238\n",
            "--- 4108.41557765007 seconds ---\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DsqmAYO1khSW",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "### Printing Prediction for Submission on Kaggle\n",
        "root = f_path = '/content/drive/My Drive/Team 3 Kaggle Competition ML2/'\n",
        "catDict = {\n",
        "  \"Women\": 0,\n",
        "  \"Men\": 1,\n",
        "  \"Girls\": 2,\n",
        "  \"Boys\": 3,\n",
        "  \"Baby\": 4\n",
        "}\n",
        "\n",
        "def print_predictions(representation,classifier):\n",
        "  input_representation = representation\n",
        "  predictions = open(root + \"prediction_Category_\" + input_representation + '_'+ classifier + \".txt\", 'w')\n",
        "  predictions.write(\"reviewerID-reviewHash,category\\n\")\n",
        "  for l in read_json(root + \"test_Category.json\"):\n",
        "    cat = 'Women' # Default category is the most popular category\n",
        "    input_text = l['reviewText'].lower()\n",
        "    input_text = clean_text(input_text)\n",
        "    if input_representation == 'tfidf':\n",
        "      label = tfidf.predict([input_text])\n",
        "      cat = catDict[label[0]]\n",
        "    if input_representation == 'ngram_tfidf':\n",
        "      input_text = ngram.transform([input_text])\n",
        "      label = ngram_tfidf.predict(input_text)\n",
        "      cat = catDict[label[0]]\n",
        "    if input_representation == 'biLSTM':\n",
        "      tokenizer.fit_on_texts([input_text])\n",
        "      sequences = tokenizer.texts_to_sequences([input_text])\n",
        "      input_text = pad_sequences(sequences, maxlen=max_len)\n",
        "      vec_predict = model.predict(input_text)\n",
        "      label = [np.argmax(vec_predict)]\n",
        "      cat = label[0]\n",
        "\n",
        "\n",
        "    # print(label)\n",
        "    # print(cat)\n",
        "    predictions.write(l['reviewerID'] + '-' + l['reviewHash'] + ',' + str(cat) + \"\\n\")\n",
        "\n",
        "  predictions.close()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PJz_0zRMCu5F",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# print_predictions('tfidf', 'logistic_regression')\n",
        "# print_predictions('ngram_tfidf', 'logistic_regression')\n",
        "print_predictions('biLSTM', 'logistic_regression')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PlWWlTLH-Hcu",
        "colab_type": "text"
      },
      "source": [
        "# New Section"
      ]
    }
  ]
}