Authorship_classification.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Authorship classification.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "RjmJ0x0Nky_B",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "0fc2c8a6-083b-45bf-c1b6-ccfd8e8f64ae"
      },
      "source": [
        "w!pip install -U -q PyDrive\n",
        "from pydrive.auth import GoogleAuth\n",
        "from pydrive.drive import GoogleDrive\n",
        "from google.colab import auth\n",
        "from oauth2client.client import GoogleCredentials\n",
        "import warnings\n",
        "warnings.filterwarnings('ignore')\n",
        "\n",
        "# Authenticate and create the PyDrive client.\n",
        "auth.authenticate_user()\n",
        "gauth = GoogleAuth()\n",
        "gauth.credentials = GoogleCredentials.get_application_default()\n",
        "drive = GoogleDrive(gauth)\n",
        "\n",
        "# The dataset shareable link\n",
        "\n",
        "#new stylo data\n",
        "#link = 'https://drive.google.com/open?id=15e0XWZR1CYx0nyBOLasWJuwMyFgF-U5l'\n",
        "link = 'https://drive.google.com/file/d/1LEbg13-KBZzHmIWAv9SnthGUzjvcbJuA/'\n",
        "id= '1LEbg13-KBZzHmIWAv9SnthGUzjvcbJuA'\n",
        "\n",
        "#fluff, id = link.split('=')\n",
        "filename = 'stylo-data.csv'\n",
        "downloaded = drive.CreateFile({'id':id})\n",
        "downloaded.GetContentFile(filename)\n",
        "\n",
        "\n",
        "#raw_data_link = 'https://drive.google.com/file/d/1DZkGVcBNQ1WUk11dVRzJcbGNeAb8rOaJ/view?usp=sharing'\n",
        "raw_id = '1DZkGVcBNQ1WUk11dVRzJcbGNeAb8rOaJ'\n",
        "#fluff, id = raw_data_link.split('=')\n",
        "print(raw_id)\n",
        "raw_filename = 'raw_data.csv'\n",
        "downloaded = drive.CreateFile({'id':raw_id})\n",
        "downloaded.GetContentFile(raw_filename)\n",
        "\n",
        "\n",
        "#news stylo dataset\n",
        "news_id = '15zZXsks6cI0FY4vpL555lBQSBr1QyjT4'\n",
        "news_filename = 'news-data.csv'\n",
        "downloaded = drive.CreateFile({'id':news_id})\n",
        "downloaded.GetContentFile(news_filename)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "1DZkGVcBNQ1WUk11dVRzJcbGNeAb8rOaJ\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QDJ9k4ISEyG2"
      },
      "source": [
        "import pandas as pd\n",
        "\n",
        "import numpy as np\n",
        "from matplotlib import pyplot as plt\n",
        "\n",
        "import nltk\n",
        "nltk.download('stopwords')\n",
        "nltk.download('punkt')\n",
        "\n",
        "from sklearn.model_selection import train_test_split, cross_validate, StratifiedKFold, RandomizedSearchCV\n",
        "from sklearn.metrics import classification_report, confusion_matrix, accuracy_score\n",
        "from sklearn.preprocessing import LabelEncoder, StandardScaler\n",
        "from sklearn.mixture import GaussianMixture, BayesianGaussianMixture\n",
        "from sklearn.cluster import KMeans\n",
        "from sklearn.decomposition import PCA\n",
        "from sklearn.pipeline import Pipeline, make_pipeline\n",
        "from sklearn.linear_model import SGDClassifier, LogisticRegression\n",
        "from sklearn.ensemble import AdaBoostClassifier, BaggingClassifier, ExtraTreesClassifier, GradientBoostingClassifier, RandomForestClassifier\n",
        "from sklearn.svm import SVC, LinearSVC\n",
        "from sklearn.feature_extraction.text import CountVectorizer, TfidfTransformer,TfidfVectorizer\n",
        "from sklearn.naive_bayes import GaussianNB, BernoulliNB, MultinomialNB\n",
        "from sklearn.neural_network import MLPClassifier\n",
        "from sklearn.neighbors import KNeighborsClassifier\n",
        "from sklearn.feature_selection import SelectKBest, VarianceThreshold, SelectPercentile, SelectFwe, RFECV,  chi2, f_classif, mutual_info_classif\n",
        "from sklearn.preprocessing import StandardScaler, MinMaxScaler, PowerTransformer, QuantileTransformer, Normalizer\n",
        "\n",
        "!pip install -U keras\n",
        "!pip install -U shap\n",
        "!pip install -U eli5\n",
        "import shap\n",
        "import eli5\n",
        "\n",
        "from keras.utils import np_utils\n",
        "from keras.preprocessing.text import Tokenizer\n",
        "from keras.preprocessing.sequence import pad_sequences\n",
        "from tensorflow.python.keras.layers import InputLayer, Embedding, Conv1D, MaxPool1D, Dense, Flatten, Dropout, GRU, LSTM, Input\n",
        "from tensorflow.python.keras import Sequential, Model\n",
        "from tensorflow.python.keras import Sequential\n",
        "from keras.initializers import Constant\n",
        "\n",
        "from pprint import pprint\n",
        "\n",
        "from imblearn.over_sampling import SMOTE"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pbymYhVUjAYD",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "706458bd-228d-450a-8055-6fdc6efdb0fc"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Mounted at /content/drive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "UdSHLajCxFOr"
      },
      "source": [
        "*PS: Para baixar os arquivos do Google Drive, é preciso clicar no link que irá aparecer no console, aceitar as permissões do Drive e em seguida colar o código de verificação no local indicado (console)*\n",
        "\n",
        "*Sem o arquivo não é possível executar os outros trechos do notebook*"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sZ9Qh3LPNDN_"
      },
      "source": [
        "# read dataset\n",
        "df = pd.read_csv(filename)\n",
        "raw_df = pd.read_csv(raw_filename)\n",
        "df_news = pd.read_csv(news_filename)\n",
        "\n",
        "\n",
        "# Remove single class entries (We need at least 3 samples of an author)#\n",
        "df = df.groupby('Author').filter(lambda x: len(x) > 3)\n",
        "raw_df = raw_df.groupby('Author').filter(lambda x: len(x) > 3)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "B-SGc_MBRjyN"
      },
      "source": [
        "#Separate data from labels\n",
        "\n",
        "y = df.pop('Author')\n",
        "y_raw = raw_df.pop('Author')\n",
        "y_news = df_news.pop('Author')\n",
        "\n",
        "le = LabelEncoder()\n",
        "le.fit(y)\n",
        "encoded_Y = le.transform(y)\n",
        "\n",
        "le.fit(y_raw)\n",
        "encoded_raw_Y = le.transform(y_raw)\n",
        "\n",
        "le.fit(y_news)\n",
        "encoded_y_news = le.transform(y_news)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yYSCvf-g1oyd"
      },
      "source": [
        "raw_df"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SsqfbzdbtMfV",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "6bc13107-b246-453f-dc79-06eae9f3168e"
      },
      "source": [
        "# Feature selection\n",
        "\n",
        "# Baseline (Variance threshold):\n",
        "threshold=(.01)\n",
        "selector = VarianceThreshold(threshold)\n",
        "selector.fit(df)\n",
        "\n",
        "df.columns[selector.get_support(indices=True)] # 22 features, data has very low variance (features are mostly very close)\n",
        "\n",
        "# selector.fit(df)\n",
        "\n",
        "# df.columns[selector.get_support(indices=True)]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "22"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 52
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "R1dCWlly5who"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_qFPfnjikFk7",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "b7304dd1-c842-4c46-df4c-10fc6552ce8f"
      },
      "source": [
        "# 50% better based on percentile using chi2 (37f)\n",
        "selector = SelectPercentile(chi2, percentile=50)\n",
        "selector.fit(df, y)\n",
        "df.columns[selector.get_support(indices=True)]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Index(['DiversidadeLexica', 'TamanhoMedioDasPalavras', 'TamanhoMedioSentencas',\n",
              "       'StdevSentencas', 'TamanhoMedioParagrafos', 'StdevTamParagrafos',\n",
              "       'FrequenciaDeParagrafos', 'Monossilabas', 'Ponto', 'Virgulas',\n",
              "       'Exclamacoes', 'DoisPontos', 'Citacoes', 'QuebrasDeLinha', 'Digitos',\n",
              "       'PronomesPorPreposicao', 'Passado', 'EntAbstracao', 'EntAcontecimento',\n",
              "       'EntCoisa', 'EntLocal', 'EntOrganizacao', 'EntObra', 'EntOutro',\n",
              "       'EntPessoa', 'EntTempo', 'EntValor', 'GuiraudR', 'HerdanV', 'DugastU',\n",
              "       'HonoresH', 'HapaxLegomenaLocal', 'BRFleshIndex', 'top 1-gram',\n",
              "       'top 3-gram', 'top 4-gram', 'top 5-gram'],\n",
              "      dtype='object')"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4TlAWkNb3s6W",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "dc730221-e7e0-46d2-ed5c-43322415bcc7"
      },
      "source": [
        "# 50% better based on percentile using chi2 (37f)\n",
        "selector = SelectPercentile(chi2, percentile=50)\n",
        "selector.fit(df_news, y_news)\n",
        "df.columns[selector.get_support(indices=True)]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Index(['DiversidadeLexica', 'TamanhoMedioDasPalavras', 'TamanhoMedioSentencas',\n",
              "       'StdevSentencas', 'TamanhoMedioParagrafos', 'StdevTamParagrafos',\n",
              "       'Monossilabas', 'Ponto', 'Virgulas', 'Exclamacoes', 'DoisPontos',\n",
              "       'Citacoes', 'QuebrasDeLinha', 'Adverbios', 'VerbosPtcp', 'Pronomes',\n",
              "       'PronomesPorPreposicao', 'TermosNaoTageados', 'PrimeiraPessoa',\n",
              "       'Presente', 'EntCoisa', 'EntObra', 'EntOutro', 'EntPessoa', 'EntTempo',\n",
              "       'EntValor', 'GuiraudR', 'HerdanV', 'DugastU', 'PalavrasErroOrtografico',\n",
              "       'HapaxLegomenaLocal', 'BRFleshIndex', 'PalavrasCapitalizadas',\n",
              "       'top 1-gram', 'top 3-gram', 'top 4-gram', 'top 5-gram'],\n",
              "      dtype='object')"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 9
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cbKN8D8OJtkY",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "14326be8-d427-42f4-f28a-7f85b13e72c9"
      },
      "source": [
        "selector = SelectPercentile(percentile=50)\n",
        "selector.fit(df, y)\n",
        "df.columns[selector.get_support(indices=True)]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Index(['DiversidadeLexica', 'TamanhoMedioDasPalavras', 'TamanhoMedioSentencas',\n",
              "       'StdevSentencas', 'TamanhoMedioParagrafos', 'FrequenciaDeParagrafos',\n",
              "       'FrequenciaPalavrasDuplicadas', 'MediaSilabasPorPalavra',\n",
              "       'Monossilabas', 'Ponto', 'Virgulas', 'Citacoes', 'QuebrasDeLinha',\n",
              "       'Digitos', 'Adjetivos', 'Preposicoes', 'Conjuncoes',\n",
              "       'TermosNaoTageados', 'PalavrasDeConteudo', 'GenMasc', 'GenFem',\n",
              "       'SemGenero', 'TotalEntidadesNomeadas', 'HerdanC', 'DugastU', 'MaasA',\n",
              "       'HonoresH', 'PalavrasErroOrtografico', 'HapaxLegomenaLocal',\n",
              "       'PalavrasComunsTam2', 'StopWords', 'BRFleshIndex', 'Global hapax',\n",
              "       'top 1-gram', 'top 3-gram', 'top 4-gram', 'top 5-gram'],\n",
              "      dtype='object')"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TncVnkdo4anr",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "7ce88148-9df2-44bb-efed-5ee35fb17b4b"
      },
      "source": [
        "selector = SelectPercentile(percentile=50)\n",
        "selector.fit(df_news, y_news)\n",
        "df_news.columns[selector.get_support(indices=True)]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Index(['DiversidadeLexica', 'TamanhoMedioDasPalavras', 'StdevSentencas',\n",
              "       'TamanhoMedioParagrafos', 'StdevTamParagrafos',\n",
              "       'FrequenciaDeParagrafos', 'Monossilabas', 'Exclamacoes', 'Citacoes',\n",
              "       'QuebrasDeLinha', 'Adverbios', 'Substantivos', 'Preposicoes',\n",
              "       'VerbosPtcp', 'Pronomes', 'PronomesPorPreposicao', 'TermosNaoTageados',\n",
              "       'PalavrasDeConteudo', 'FrasesVerbais', 'SemGenero', 'Singular',\n",
              "       'TerceiraPessoa', 'Presente', 'Futuro', 'GuiraudR', 'HerdanV',\n",
              "       'MedidaK', 'DugastU', 'MaasA', 'PalavrasErroOrtografico',\n",
              "       'HapaxLegomenaLocal', 'PalavrasComunsTam4', 'StopWords',\n",
              "       'PalavrasCapitalizadas', 'Global hapax', 'top 4-gram', 'top 5-gram'],\n",
              "      dtype='object')"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 12
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bs8jp-5XxMC8",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "3643d69a-516b-4fea-f72b-c96f34241367"
      },
      "source": [
        "# K-Best (chi2)\n",
        "best_k_features = SelectKBest(chi2, k=74).fit(df, y)\n",
        "dfscores = pd.DataFrame(best_k_features.scores_)\n",
        "dfcolumns = pd.DataFrame(df.columns)\n",
        "featureScores = pd.concat([dfcolumns,dfscores],axis=1)\n",
        "featureScores.columns = ['Feature','Score'] \n",
        "print(featureScores.nlargest(35,'Score'))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "                    Feature        Score\n",
            "55                  HerdanV  3240.480483\n",
            "4    TamanhoMedioParagrafos  2660.335980\n",
            "5        StdevTamParagrafos  1373.072039\n",
            "3            StdevSentencas   150.375319\n",
            "0         DiversidadeLexica   138.597858\n",
            "2     TamanhoMedioSentencas   126.791038\n",
            "70               top 1-gram   114.339214\n",
            "57                  DugastU    79.869125\n",
            "13               DoisPontos    55.254317\n",
            "11                 Virgulas    35.802381\n",
            "15           QuebrasDeLinha    25.352158\n",
            "12              Exclamacoes    25.096919\n",
            "10                    Ponto    17.111885\n",
            "53                 GuiraudR    15.582640\n",
            "66             BRFleshIndex     9.122961\n",
            "71               top 3-gram     6.813025\n",
            "26    PronomesPorPreposicao     5.663002\n",
            "72               top 4-gram     4.827784\n",
            "73               top 5-gram     4.433602\n",
            "1   TamanhoMedioDasPalavras     2.611150\n",
            "59                 HonoresH     2.088297\n",
            "61       HapaxLegomenaLocal     1.980219\n",
            "44         EntAcontecimento     1.915102\n",
            "49                 EntOutro     1.436834\n",
            "43             EntAbstracao     1.397091\n",
            "47           EntOrganizacao     1.258447\n",
            "51                 EntTempo     1.088513\n",
            "52                 EntValor     1.085980\n",
            "6    FrequenciaDeParagrafos     1.077922\n",
            "48                  EntObra     1.064877\n",
            "45                 EntCoisa     1.043115\n",
            "16                  Digitos     1.039257\n",
            "50                EntPessoa     1.022828\n",
            "9              Monossilabas     0.926692\n",
            "46                 EntLocal     0.825038\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eA6hAlJi5ARi",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "63a8b1d4-dfa6-4847-fe45-619bcb087c48"
      },
      "source": [
        "# K-Best (chi2)\n",
        "best_k_features = SelectKBest(chi2, k=74).fit(df_news, y_news)\n",
        "dfscores = pd.DataFrame(best_k_features.scores_)\n",
        "dfcolumns = pd.DataFrame(df_news.columns)\n",
        "featureScores = pd.concat([dfcolumns,dfscores],axis=1)\n",
        "featureScores.columns = ['Feature','Score'] \n",
        "print(featureScores.nlargest(35,'Score'))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "                    Feature        Score\n",
            "55                  HerdanV  2688.503948\n",
            "5        StdevTamParagrafos  1707.831479\n",
            "4    TamanhoMedioParagrafos  1103.327474\n",
            "2     TamanhoMedioSentencas   862.748893\n",
            "12              Exclamacoes    59.697231\n",
            "3            StdevSentencas    59.007230\n",
            "0         DiversidadeLexica    37.698913\n",
            "57                  DugastU    33.392403\n",
            "53                 GuiraudR    22.573746\n",
            "11                 Virgulas    21.582665\n",
            "10                    Ponto    16.308883\n",
            "13               DoisPontos    12.164414\n",
            "70               top 1-gram    10.241280\n",
            "26    PronomesPorPreposicao     6.568906\n",
            "66             BRFleshIndex     4.978543\n",
            "72               top 4-gram     2.430000\n",
            "73               top 5-gram     2.430000\n",
            "1   TamanhoMedioDasPalavras     1.791193\n",
            "68    PalavrasCapitalizadas     0.921331\n",
            "27        TermosNaoTageados     0.810441\n",
            "71               top 3-gram     0.713333\n",
            "37           PrimeiraPessoa     0.702100\n",
            "9              Monossilabas     0.686709\n",
            "49                 EntOutro     0.551372\n",
            "48                  EntObra     0.526174\n",
            "25                 Pronomes     0.487441\n",
            "61       HapaxLegomenaLocal     0.448121\n",
            "60  PalavrasErroOrtografico     0.419544\n",
            "14                 Citacoes     0.399479\n",
            "45                 EntCoisa     0.389917\n",
            "50                EntPessoa     0.377552\n",
            "15           QuebrasDeLinha     0.358533\n",
            "51                 EntTempo     0.349568\n",
            "18                Adverbios     0.333384\n",
            "23               VerbosPtcp     0.316366\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RpEw6KTq27Q6",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "7531af17-a560-4d4d-baae-52ab7cd278d9"
      },
      "source": [
        "# K-Best (f_classif)\n",
        "best_k_features = SelectKBest(f_classif, k=54).fit(df_news, y_news)\n",
        "dfscores = pd.DataFrame(best_k_features.scores_)\n",
        "dfcolumns = pd.DataFrame(df_news.columns)\n",
        "featureScores = pd.concat([dfcolumns,dfscores],axis=1)\n",
        "featureScores.columns = ['Feature','Score'] \n",
        "print(featureScores.nlargest(35,'Score'))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "                    Feature      Score\n",
            "72               top 4-gram  81.000000\n",
            "73               top 5-gram  81.000000\n",
            "12              Exclamacoes  19.089614\n",
            "69             Global hapax  13.503812\n",
            "53                 GuiraudR  13.162995\n",
            "25                 Pronomes  12.735249\n",
            "28       PalavrasDeConteudo  12.442170\n",
            "27        TermosNaoTageados  12.376033\n",
            "56                  MedidaK  12.152812\n",
            "26    PronomesPorPreposicao  10.844617\n",
            "1   TamanhoMedioDasPalavras  10.182217\n",
            "18                Adverbios   9.488145\n",
            "9              Monossilabas   9.470314\n",
            "65                StopWords   9.293226\n",
            "55                  HerdanV   9.033462\n",
            "68    PalavrasCapitalizadas   8.862075\n",
            "21              Preposicoes   8.457826\n",
            "20             Substantivos   8.408751\n",
            "6    FrequenciaDeParagrafos   8.297498\n",
            "40                 Presente   7.427606\n",
            "15           QuebrasDeLinha   7.420883\n",
            "23               VerbosPtcp   6.958381\n",
            "4    TamanhoMedioParagrafos   6.232976\n",
            "38           TerceiraPessoa   6.224789\n",
            "57                  DugastU   6.214147\n",
            "60  PalavrasErroOrtografico   6.084080\n",
            "0         DiversidadeLexica   5.951390\n",
            "35                 Singular   5.758693\n",
            "58                    MaasA   5.555713\n",
            "34                SemGenero   5.453416\n",
            "41                   Futuro   5.447073\n",
            "5        StdevTamParagrafos   5.414068\n",
            "31            FrasesVerbais   5.323598\n",
            "14                 Citacoes   5.060375\n",
            "64       PalavrasComunsTam4   4.877481\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dtJLynQx3Cfo",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "371d03b4-604c-4841-9b57-8f3ced2c9442"
      },
      "source": [
        "from sklearn.ensemble import ExtraTreesClassifier\n",
        "from sklearn.feature_selection import SelectFromModel\n",
        "\n",
        "#SelectFromModel will select those features which importance is greater than the mean importance of all the features by default\n",
        "sel = SelectFromModel(ExtraTreesClassifier(n_estimators = 200, random_state=0))\n",
        "sel.fit(df_news, y_news)\n",
        "selected_feat= df.columns[(sel.get_support())]\n",
        "print(len(selected_feat))\n",
        "print(selected_feat)\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "30\n",
            "Index(['TamanhoMedioDasPalavras', 'TamanhoMedioSentencas', 'StdevSentencas',\n",
            "       'TamanhoMedioParagrafos', 'StdevTamParagrafos',\n",
            "       'FrequenciaDeParagrafos', 'Monossilabas', 'Ponto', 'Virgulas',\n",
            "       'Exclamacoes', 'DoisPontos', 'QuebrasDeLinha', 'Adverbios',\n",
            "       'Preposicoes', 'Pronomes', 'TermosNaoTageados', 'PalavrasDeConteudo',\n",
            "       'GuiraudR', 'HerdanV', 'MedidaK', 'DugastU', 'MaasA',\n",
            "       'PalavrasComunsTam3', 'PalavrasComunsTam4', 'StopWords',\n",
            "       'PalavrasCapitalizadas', 'Global hapax', 'top 1-gram', 'top 4-gram',\n",
            "       'top 5-gram'],\n",
            "      dtype='object')\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "v1zOqpR9PALf",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 561
        },
        "outputId": "0e20e76b-6972-40bb-f401-1755db2a8f42"
      },
      "source": [
        "clf = ExtraTreesClassifier(n_estimators = 200, random_state=0)\n",
        "clf.fit(df, y)\n",
        "#Pass the model instance along with the feature names to ELI5\n",
        "eli5.show_weights(clf, feature_names=list(df.columns), top = 30)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "\n",
              "    <style>\n",
              "    table.eli5-weights tr:hover {\n",
              "        filter: brightness(85%);\n",
              "    }\n",
              "</style>\n",
              "\n",
              "\n",
              "\n",
              "    \n",
              "\n",
              "    \n",
              "\n",
              "    \n",
              "\n",
              "    \n",
              "\n",
              "    \n",
              "\n",
              "    \n",
              "\n",
              "\n",
              "    \n",
              "\n",
              "    \n",
              "\n",
              "    \n",
              "\n",
              "    \n",
              "\n",
              "    \n",
              "\n",
              "    \n",
              "\n",
              "\n",
              "    \n",
              "\n",
              "    \n",
              "\n",
              "    \n",
              "\n",
              "    \n",
              "\n",
              "    \n",
              "        <table class=\"eli5-weights eli5-feature-importances\" style=\"border-collapse: collapse; border: none; margin-top: 0em; table-layout: auto;\">\n",
              "    <thead>\n",
              "    <tr style=\"border: none;\">\n",
              "        <th style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">Weight</th>\n",
              "        <th style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">Feature</th>\n",
              "    </tr>\n",
              "    </thead>\n",
              "    <tbody>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 80.00%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0217\n",
              "                \n",
              "                    &plusmn; 0.0497\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                HerdanC\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 80.38%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0211\n",
              "                \n",
              "                    &plusmn; 0.0457\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                Virgulas\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 81.13%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0199\n",
              "                \n",
              "                    &plusmn; 0.0430\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                Digitos\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 81.29%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0197\n",
              "                \n",
              "                    &plusmn; 0.0449\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                StopWords\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 81.38%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0196\n",
              "                \n",
              "                    &plusmn; 0.0417\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                StdevSentencas\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 81.97%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0187\n",
              "                \n",
              "                    &plusmn; 0.0400\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                Artigos\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 82.14%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0184\n",
              "                \n",
              "                    &plusmn; 0.0419\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                DugastU\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 82.26%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0183\n",
              "                \n",
              "                    &plusmn; 0.0428\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                MaasA\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 82.49%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0179\n",
              "                \n",
              "                    &plusmn; 0.0446\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                HonoresH\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 82.66%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0177\n",
              "                \n",
              "                    &plusmn; 0.0375\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                DiversidadeLexica\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 82.72%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0176\n",
              "                \n",
              "                    &plusmn; 0.0395\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                FrasesNominais\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 82.72%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0176\n",
              "                \n",
              "                    &plusmn; 0.0450\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                EntOutro\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 82.85%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0174\n",
              "                \n",
              "                    &plusmn; 0.0385\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                TamanhoMedioParagrafos\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 83.40%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0166\n",
              "                \n",
              "                    &plusmn; 0.0398\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                FrequenciaPalavrasDuplicadas\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 83.57%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0164\n",
              "                \n",
              "                    &plusmn; 0.0402\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                Verbos\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 83.62%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0163\n",
              "                \n",
              "                    &plusmn; 0.0388\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                TamanhoMedioDasPalavras\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 83.96%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0158\n",
              "                \n",
              "                    &plusmn; 0.0374\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                HapaxLegomenaLocal\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 84.05%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0157\n",
              "                \n",
              "                    &plusmn; 0.0372\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                Global hapax\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 84.16%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0155\n",
              "                \n",
              "                    &plusmn; 0.0380\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                Preposicoes\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 84.24%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0154\n",
              "                \n",
              "                    &plusmn; 0.0359\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                Futuro\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 84.64%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0149\n",
              "                \n",
              "                    &plusmn; 0.0372\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                FrequenciaDeParagrafos\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 84.64%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0149\n",
              "                \n",
              "                    &plusmn; 0.0366\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                MedidaK\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 84.70%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0148\n",
              "                \n",
              "                    &plusmn; 0.0328\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                GuiraudR\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 84.81%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0146\n",
              "                \n",
              "                    &plusmn; 0.0344\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                GenFem\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 84.90%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0145\n",
              "                \n",
              "                    &plusmn; 0.0355\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                HerdanV\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 85.04%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0143\n",
              "                \n",
              "                    &plusmn; 0.0326\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                PalavrasFuncionais\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 85.09%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0142\n",
              "                \n",
              "                    &plusmn; 0.0341\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                TerceiraPessoa\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 85.17%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0141\n",
              "                \n",
              "                    &plusmn; 0.0355\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                Ponto\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 85.22%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0141\n",
              "                \n",
              "                    &plusmn; 0.0311\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                Conjuncoes\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "        <tr style=\"background-color: hsl(120, 100.00%, 85.22%); border: none;\">\n",
              "            <td style=\"padding: 0 1em 0 0.5em; text-align: right; border: none;\">\n",
              "                0.0141\n",
              "                \n",
              "                    &plusmn; 0.0347\n",
              "                \n",
              "            </td>\n",
              "            <td style=\"padding: 0 0.5em 0 0.5em; text-align: left; border: none;\">\n",
              "                GenMasc\n",
              "            </td>\n",
              "        </tr>\n",
              "    \n",
              "    \n",
              "        \n",
              "            <tr style=\"background-color: hsl(120, 100.00%, 85.22%); border: none;\">\n",
              "                <td colspan=\"2\" style=\"padding: 0 0.5em 0 0.5em; text-align: center; border: none; white-space: nowrap;\">\n",
              "                    <i>&hellip; 44 more &hellip;</i>\n",
              "                </td>\n",
              "            </tr>\n",
              "        \n",
              "    \n",
              "    </tbody>\n",
              "</table>\n",
              "    \n",
              "\n",
              "    \n",
              "\n",
              "\n",
              "    \n",
              "\n",
              "    \n",
              "\n",
              "    \n",
              "\n",
              "    \n",
              "\n",
              "    \n",
              "\n",
              "    \n",
              "\n",
              "\n",
              "\n"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 77
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "VoyMWASA6d0W",
        "outputId": "ee2e6aac-8675-4641-8b48-b1cd3f27e7c4"
      },
      "source": [
        "encoded_y_news"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 1, 1,\n",
              "       1, 1, 1, 1, 1, 1, 1, 1, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 5, 5, 5, 5,\n",
              "       5, 5, 5, 5, 5, 5, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 6, 6, 6, 6, 6, 6,\n",
              "       6, 6, 6, 6, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0,\n",
              "       0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2])"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ifIFpGFaSsJC",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 581
        },
        "outputId": "9e24b164-89e9-40a6-9c62-a7c9a3541ee6"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 576x684 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QHrOSn3i64ot",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 581
        },
        "outputId": "9160138c-9c8c-4b6f-df24-68a890d1f0af"
      },
      "source": [
        "# Create the RFE object and compute a cross-validated score.\n",
        "rfc = RandomForestClassifier(n_estimators=200)\n",
        "# The \"accuracy\" scoring is proportional to the number of correct classifications\n",
        "\n",
        "rfecv = RFECV(estimator=rfc, step=1, cv=5,scoring='accuracy')\n",
        "rfecv.fit(df_news, y_news)\n",
        "\n",
        "print(\"Optimal number of features news-dataset: %d\" % rfecv.n_features_)\n",
        "\n",
        "rfecv2 = RFECV(estimator=rfc, step=1, cv=3,\n",
        "              scoring='accuracy')\n",
        "rfecv2.fit(df, y)\n",
        "\n",
        "print(\"Optimal number of features students-dataset: %d\" % rfecv2.n_features_)\n",
        "\n",
        "fig, (ax1, ax2) = plt.subplots(1, 2)\n",
        "fig.set_size_inches(12, 8)\n",
        "\n",
        "ax1.plot(range(1, len(rfecv.grid_scores_) + 1), rfecv.grid_scores_)\n",
        "ax1.set_xlabel(\"Number of features selected\")\n",
        "ax1.set_ylabel(\"Cross validation score (nb of correct classifications)\")\n",
        "ax1.set_title('Random Forest n=200 - News dataset')\n",
        "\n",
        "ax2.plot(range(1, len(rfecv2.grid_scores_) + 1), rfecv2.grid_scores_)\n",
        "ax2.set_xlabel(\"Number of features selected\")\n",
        "ax2.set_ylabel(\"Cross validation score (nb of correct classifications)\")\n",
        "ax2.set_title('Random Forest n=200 - Students dataset')\n",
        "\n",
        "plt.figure()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Optimal number of features news-dataset: 71\n",
            "Optimal number of features students-dataset: 50\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 0 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 864x576 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 0 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SrMzJ2vOtsZG",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 316
        },
        "outputId": "cff3f05b-67b7-4495-fc2c-22f0eb8c9cc6"
      },
      "source": [
        "svc = RandomForestClassifier(n_estimators=200)\n",
        "# The \"accuracy\" scoring is proportional to the number of correct\n",
        "# classifications\n",
        "rfecv = RFECV(estimator=svc, step=1, cv=5,\n",
        "              scoring='accuracy')\n",
        "rfecv.fit(df_news, y_news)\n",
        "\n",
        "print(\"Optimal number of features : %d\" % rfecv.n_features_)\n",
        "\n",
        "# Plot number of features VS. cross-validation scores\n",
        "plt.figure()\n",
        "plt.xlabel(\"Number of features selected\")\n",
        "plt.ylabel(\"Cross validation score (nb of correct classifications)\")\n",
        "plt.plot(range(1, len(rfecv.grid_scores_) + 1), rfecv.grid_scores_)\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Optimal number of features : 34\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xX-rbG4pTw8R",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "outputId": "5590747d-290b-44f4-aa71-6b6c9f5ef52e"
      },
      "source": [
        "# Visualize data\n",
        "df.shape\n",
        "df.head()\n",
        "df.shape\n",
        "raw_df.head()"
      ],
      "execution_count": null,
      "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>Text</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>respostas ecologia evolutiva retirar amostra p...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>1 - competição membros espécies diferentes usa...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>embora temática questão ambiental crescente an...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5</th>\n",
              "      <td>vem cada dia buscando entender relações estabe...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>6</th>\n",
              "      <td>pode-se dizer que, conservar dado habitat dada...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                                                Text\n",
              "1  respostas ecologia evolutiva retirar amostra p...\n",
              "2  1 - competição membros espécies diferentes usa...\n",
              "3  embora temática questão ambiental crescente an...\n",
              "5  vem cada dia buscando entender relações estabe...\n",
              "6  pode-se dizer que, conservar dado habitat dada..."
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 18
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EWwBfLgS-zZA"
      },
      "source": [
        "#stopwords = nltk.corpus.stopwords.words('portuguese')\n",
        "raw_df['Text'] = raw_df['Text'].apply(lambda x: ' '.join([word for word in x.split()]))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gxmBkJBicT8u"
      },
      "source": [
        "raw_df = raw_df.replace(to_replace ='\\r', value = ' ', regex = True) \n",
        "raw_df = raw_df.replace(to_replace ='\\n ', value = '\\n', regex = True) \n",
        "raw_df = raw_df.replace(to_replace ='\\n{2,*}', value = ' ', regex = True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2zKovLDZF0Le",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "8ae00653-9491-4fd9-c5fa-4da8c1524adb"
      },
      "source": [
        "raw_df['Text']"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "1      respostas ecologia evolutiva retirar uma amost...\n",
              "2      1 - competição entre membros de espécies difer...\n",
              "3      embora a temática questão ambiental esteja num...\n",
              "5      a ciência vem a cada dia buscando entender as ...\n",
              "6      pode-se dizer que, conservar um dado habitat o...\n",
              "                             ...                        \n",
              "92     referencial teórico: anuras ( bufonídeos ) ou ...\n",
              "93     congregação santa dorotéia do brasil – – a seg...\n",
              "94     congregação santa doroteia do brasil em ciênci...\n",
              "95     disciplina: orientação e supervisão de estágio...\n",
              "102    metodologia do ensino de ciências e de biologi...\n",
              "Name: Text, Length: 69, dtype: object"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 11
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XKDn9mh9-DJ7"
      },
      "source": [
        "#X_idf = tfidf_transformer.fit_transform(raw_df['Text'])\n",
        "X_tfidf = TfidfVectorizer().fit_transform(raw_df['Text'])\n",
        "\n",
        "from sklearn.decomposition import TruncatedSVD\n",
        "\n",
        "svd = TruncatedSVD(n_components=3, random_state=42)\n",
        "data = svd.fit_transform(X_tfidf) "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MOMw5Fol0K4e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 280
        },
        "outputId": "41899cbd-3a1a-4246-d9ee-8c5a910412b3"
      },
      "source": [
        "gx = GaussianMixture()\n",
        "\n",
        "#gx.fit(df)\n",
        "n_components = np.arange(1, 30)\n",
        "models = [GaussianMixture(n, covariance_type='full', random_state=0).fit(df) for n in n_components]\n",
        "plt.plot(n_components, [m.bic(df) for m in models], label='BIC')\n",
        "plt.plot(n_components, [m.aic(df) for m in models], label='AIC')\n",
        "plt.legend(loc='best')\n",
        "plt.xlabel('n_components');"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEHCAYAAABfkmooAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAAgAElEQVR4nO3deXhU5fXA8e8hAcIWlrATQsK+yRrZVKS4gRvuYq3iirbuv7pga93bam1r0VpbLCq0FkQWQUWRTUWRLWwJCUsMSwIhCQQSQkjIcn5/3BsJCASy3czM+TzPPDPz3jsz5zo4J/e+73teUVWMMcaY06nldQDGGGNqPksWxhhjymTJwhhjTJksWRhjjCmTJQtjjDFlsmRhjDGmTMFeB1DZmjdvrpGRkV6HYYwxPiUmJmafqrY41Xa/SxaRkZGsWbPG6zCMMcaniMjO0223y1DGGGPKZMnCGGNMmSxZGGOMKZPf9VmcTEFBASkpKeTl5XkdSqUICQkhPDyc2rVrex2KMSZABESySElJoVGjRkRGRiIiXodTIarK/v37SUlJISoqyutwjDEBIiAuQ+Xl5REWFubziQJARAgLC/ObsyRjjG8IiGQB+EWiKOFPx2KMqRyH8wur9P0DJll4LSgoiH79+tG3b18GDBjA8uXLAdixYwe9e/f+cb9Vq1YxfPhwunXrRv/+/bnnnnvIzc31KmxjTA2nqsxYk8x5ry4hZueBKvucgOizqAnq1avH+vXrAViwYAFPP/00X3/99XH7pKWlceONNzJ9+nSGDh0KwMyZMzl06BD169ev9piNMTVbatYRnp4dy1dbMhgU1YzmDetU2WdZsvBAdnY2TZs2/Un7W2+9xbhx435MFAA33HBDdYZmjPEBqspHa1J46dN4CouV56/qye1DI6lVq+ouUQdcsnjhk03E78mu1Pfs2TaU567qddp9jhw5Qr9+/cjLyyM1NZUlS5b8ZJ+4uDjGjRtXqbEZY/xLatYRJsyK5eutztnEazf0oUNYgyr/3IBLFl4pfRnq+++/5/bbbycuLs7jqIwxvuLEs4kXru7FbUM6VOnZRGkBlyzKOgOoDkOHDmXfvn1kZGQc196rVy9iYmIYM2aMR5EZY2qiPQedvomvt2YwOKoZf6qms4nSbDSUBzZv3kxRURFhYWHHtT/44INMmTKFlStX/tg2e/Zs0tLSqjtEY0wNoKp8uHoXl73+Dau2Z/LC1b2Ydu+Qak8UEIBnFl4p6bMA5x/AlClTCAoKOm6fVq1aMX36dB5//HHS09OpVasWw4cPZ9SoUV6EbIzxUHp2Hk/M3Ojp2URpZSYLEXkXuBJIV9XeJ2z7NfBnoIWq7hNntthE4HIgF7hDVde6+44DnnFf+rKqTnHbBwLvA/WA+cAjqqoi0gz4EIgEdgA3qWrVDSKuYkVFRSdtj4yMPK7vYujQoSxbtqy6wjLG1ECL4tN4ctZGco8WVnvfxKmcyWWo94Gf/GkrIu2BS4FdpZpHA13c23jgbXffZsBzwGBgEPCciJSMHX0buLfU60o+awKwWFW7AIvd58YY47eOHC3imY9juWfqGlqHhvDpQ+czbljVDok9U2UmC1X9Bsg8yabXgScBLdU2BpiqjhVAExFpA1wGLFTVTPfsYCEwyt0WqqorVFWBqcA1pd5rivt4Sql2Y4zxOwmp2Vz992/574pd3HtBFHMeGEbnlo28DutH5eqzEJExwG5V3XBCnaJ2QHKp5ylu2+naU07SDtBKVVPdx3uBVqeJZzzOmQwRERFnezjGGOMZVeW973bwyuebaVy/NlPvGsTwrqdcCtszZ50sRKQ+8BucS1DVwu3D0NNsnwRMAoiOjj7lfsYYU5NkHMrn8Y828PXWDC7u0ZJXr+9DWMO6Xod1UuU5s+gERAElZxXhwFoRGQTsBtqX2jfcbdsNjDih/Su3Pfwk+wOkiUgbVU11L1ellyNWY4ypkZZuTufxjzaQk1/IS2N68YshHWp0RemznmehqrGq2lJVI1U1EufS0QBV3QvMA24XxxAgy72UtAC4VESauh3blwIL3G3ZIjLEHUl1OzDX/ah5QEnti3Gl2o0xxmflFRTx/LxN3Pn+alo0qssnD53PbUNr/sJsZSYLEZkGfA90E5EUEbn7NLvPB5KAROAd4FcAqpoJvASsdm8vum24+/zbfc0PwOdu+yvAJSKyDbjYfe7TPv74Y0SEzZs3A1ae3JhAsy3tENe89R3vL9/BnedF8vED59G1Vc3pxD6dMi9DqeotZWyPLPVYgQdOsd+7wLsnaV8D9D5J+37gorLi8yXTpk3j/PPPZ9q0abzwwgvHbbPy5Mb4L1Vl2qpkXvx0Ew3rBvPenefys24tvQ7rrFi5j2qSk5PDt99+y+TJk5k+ffpPtp+qPHmrVqccBGaM8QFZRwp48H/r+M2cWM6NbMb8Ry7wuUQBgVju4/MJsDe2ct+z9Tkw+vRXyebOncuoUaPo2rUrYWFhxMTEHFcbysqTG+N/YnZm8vC09aRl5zFhdHfGX9CxRkywKw87s6gm06ZNY+zYsQCMHTuWadOmeRyRMaaqFBUrf1+yjZv+tYJateCj+4dy/4WdfDZRQCCeWZRxBlAVMjMzWbJkCbGxsYgIRUVFiAgPPHCse8fKkxvjH9Ky83h0+nq+T9rP1X3b8vK1vQkNqe11WBVmZxbVYObMmdx2223s3LmTHTt2kJycTFRUFMnJxya1W3lyY3zf4oQ0Rv3tG9YnH+RPN/Rh4th+fpEoIBDPLDwwbdo0nnrqqeParr/+ev74xz/++NzKkxvju/ILi/jj/M28v3wHPdqE8uYt/encsqHXYVUqcUa7+o/o6Ghds2bNcW0JCQn06NHDo4iqhj8ekzG+KCkjh4emrWPTnmzuGBbJhNHdCakdVPYLaxgRiVHV6FNttzMLY4wppznrUvjtnDjqBNfindujuaSn/w51t2RhjDFn6XB+Ic/O3cSstSkMimzGxFv60aZxPa/DqlKWLIwx5iwkpGbzwP/Wsn3fYR4e2ZmHL+pCcJD/jxUKmGShqjW+UNeZ8rd+JmN8gary35W7eOnTeBrXq80Hdw9mWOfmXodVbQIiWYSEhLB//37CwsJ8PmGoKvv37yckJMTrUIwJGFlHCpgwayOfx+1leNcW/PWmvjSvoetOVJWASBbh4eGkpKSQkZHhdSiVIiQkhPDw8LJ3NMZU2LpdB3ho2jr2ZuXx9Oju3OvDJTsqIiCSRe3atYmKivI6DGOMDykuVt5ZlsRrC7bQunEIM+4fyoCIpl6H5ZmASBbGGHM2kjNzeXp2LN8m7mN079a8cn0fGtfzj5nY5XUmix+9KyLpIhJXqu01EdksIhtFZI6INCm17WkRSRSRLSJyWan2UW5boohMKNUeJSIr3fYPRaSO217XfZ7obo+srIM2xpiTKSpWJn+7nUtfd0p2/OHac/jHrQMCPlHAmdWGeh84sebEQqC3qvYBtgJPA4hIT2As0Mt9zT9EJEhEgoC3gNFAT+AWd1+AV4HXVbUzcAAoWYnvbuCA2/66u58xxlSJrWmHuP7t5bz0aTxDO4Xx5WPD+fngCJ8fFFNZykwWqvoNkHlC25eqWug+XQGU9LaOAaarar6qbsdZKnWQe0tU1SRVPQpMB8a4626PBGa6r58CXFPqvaa4j2cCF4l9a8aYSna0sJiJi7ZxxRvL2Ln/MBPH9mPyuGjaNvHvSXZnqzL6LO4CPnQft8NJHiVS3DaA5BPaBwNhwMFSiaf0/u1KXqOqhSKS5e6/rxJiNsYY1icf5KmZG9mSdogx/dry7JU9CQuwIbFnqkLJQkR+CxQCH1ROOOWOYzwwHiAiIsLLUIwxPiD3aCF//XIr7363nZaNQpg8LpqLevhvXafKUO5kISJ3AFcCF+mxKcW7gfaldgt32zhF+36giYgEu2cXpfcvea8UEQkGGrv7/4SqTgImgVN1trzHZIzxf8sT9zFhdiy7MnO5dXAEE0Z3p5GfrDlRlcpV0ERERgFPAleram6pTfOAse5IpiigC7AKWA10cUc+1cHpBJ/nJpmlwA3u68cBc0u9V8mi1DcAS9TqXBhjyimvoIjffRzHz/+9kqBawvTxQ/j9tedYojhDZZ5ZiMg0YATQXERSgOdwRj/VBRa6fc4rVPV+Vd0kIjOAeJzLUw+oapH7Pg8CC4Ag4F1V3eR+xFPAdBF5GVgHTHbbJwP/EZFEnA72sZVwvMaYALQt7RAPTVvH5r2HuOf8KB6/rJtPrjnhpYBY/MgYE5hUlQ9XJ/P8J5toUCeYv9zUlxHdWnodVo1kix8ZYwJSdl4Bv5kdy6cbUzmvcxiv39SPlqFWgLO8LFkYY/zO+uSDPDRtLXsO5vHEZd345YWdArL4X2WyZGGM8Ruli/+1Cg1hxn1DGNihmddh+QVLFsYYv5BxKJ9ff7SBb7ZmOMX/rutD4/o20qmyWLIwxvi8ZdsyeOzDDRzKK+D31/bm54OsplNls2RhjPFZ+YVF/HXhViZ9k0SnFg357z2D6N461Ouw/JIlC2OMT9qWdohHpq8nPjWbWwZF8OyVPalXx+ZOVBVLFsYYn6KqTP1+J3+Yn0CDusG8c3s0l/S0uk5VzZKFMcZnpGfn8cTMjXy9NYOfdWvBqzf0oWUjmztRHSxZGGN8wpeb9jJhdiyH8wt5aUwvfjGkg3ViVyNLFsaYGu1wfiEvfRrP9NXJ9G4Xyt9u7kfnlo28DivgWLIwxtRY63Yd4LEP17MzM5dfjujEYxd3pU5wuYplmwqyZGGMqXEKi4p5a+kPvLFkG61DQ5h27xCGdAzzOqyAZsnCGFOj7Dl4hEenr2fVjkyu6deWF8b0pnE9m4ntNUsWxpga48tNe3ly1kYKCot5/ea+XNs/3OuQjMuShTHGc3kFRfxxfgJTvt9J73ahvHnLAKKaN/A6LFNKmT1FIvKuiKSLSFyptmYislBEtrn3Td12EZE3RCRRRDaKyIBSrxnn7r9NRMaVah8oIrHua94QdyzcqT7DGONffsjI4dp/LGfK9zu567woZv1ymCWKGuhMhhW8D4w6oW0CsFhVuwCL3ecAo3HW3e4CjAfeBueHH2c51sHAIOC5Uj/+bwP3lnrdqDI+wxjjB1SVmTEpXPXmt+zNOsLkcdE8e1VP6gZbyY6aqMxkoarf4KyBXdoYYIr7eApwTan2qepYATQRkTbAZcBCVc1U1QPAQmCUuy1UVVeos77r1BPe62SfYYzxcTn5hfzfjA08/tEGzmnXmM8fGc5FPaxkR01W3j6LVqqa6j7eC5R8y+2A5FL7pbhtp2tPOUn76T7DGOPDYlOyeGjaWnZl5vLYxV15cGRngmwVuxqvwh3cqqoiopURTHk/Q0TG41z2IiIioipDMcaUk6ry7nc7eOXzBJo3rMv08UMZFGWr2PmK8k6FTHMvIeHep7vtu4H2pfYLd9tO1x5+kvbTfcZPqOokVY1W1egWLVqU85CMMVXlcH4hD/5vHS99Gs+Ibi2Z//AFlih8THmTxTygZETTOGBuqfbb3VFRQ4As91LSAuBSEWnqdmxfCixwt2WLyBB3FNTtJ7zXyT7DGONDdu4/zPVvL+fzuFQmjO7OpNsG0rRBHa/DMmepzMtQIjINGAE0F5EUnFFNrwAzRORuYCdwk7v7fOByIBHIBe4EUNVMEXkJWO3u96KqlnSa/wpnxFU94HP3xmk+wxjjI77emsHD09YB8P6dgxje1c78fZU4g5D8R3R0tK5Zs8brMIwJaKrKP79O4rUFm+naqhGTbosmIqy+12GZ0xCRGFWNPtV2m8FtjKlUuUcLeWLmRj7bmMoVfdrw2g19qF/Hfmp8nX2DxphKs2t/LuP/s4ataYeYMLo79w3vaAsU+QlLFsaYSvHN1gwesv4Jv2XJwhhTIarKv75J4k9fOP0T/7ptIB3CrLaTv7FkYYwpt4O5R/ntx3HWPxEA7Fs1xpw1VeWz2FSen7eJg7kFPDWqO/dfaP0T/syShTHmrOzNyuOZj+NYlJBGn/DGTL1rMD3bhnodlqliliyMMWekuFiZtnoXr8zfTEFxMb+9vAd3nhdJcFB5C0EYX2LJwhhTpqSMHCbMjmXV9kyGdQrjj9edY53YAcaShTHmlAqKinlnWRJ/W7SNkOBa/On6PtwYHW59EwHIkoUx5qTidmfx5MyNxKdmM7p3a164uhctQ0O8Dst4xJKFMeY4RwuL+evCrbyzLIlmDerwz18MZFTv1l6HZTxmycIY86OkjBwenr6OuN3Z3Bzdnt9c0YPG9Wp7HZapASxZGGNQVT5ak8Jz8zZRt3YtJt02kEt72dmEOcaShTEBLutIAb+ZE8tnG1MZ2jGM12/uR+vG1jdhjmfJwpgAtnpHJo9OX09adh5PjurGfcM7EVTLRjqZn6rQbBoReUxENolInIhME5EQEYkSkZUikigiH4pIHXffuu7zRHd7ZKn3edpt3yIil5VqH+W2JYrIhIrEaow5prComNcXbuXmf31PcJAw85fD+NWIzpYozCmVO1mISDvgYSBaVXsDQcBY4FXgdVXtDBwA7nZfcjdwwG1/3d0PEenpvq4XMAr4h4gEiUgQ8BYwGugJ3OLua4ypgJQDuYydtIKJi7dxTf92fPbwBfRr38TrsEwNV9F5+sFAPREJBuoDqcBIYKa7fQpwjft4jPscd/tF4szsGQNMV9V8Vd2Os373IPeWqKpJqnoUmO7ua4wpp0827GH0xGVs2XuIiWP78deb+tGwrl2NNmUr978SVd0tIn8GdgFHgC+BGOCgqha6u6UA7dzH7YBk97WFIpIFhLntK0q9denXJJ/QPri88RoTyHLyC3l+3iZmxqTQP6IJb4ztT/tmtia2OXPlThYi0hTnL/0o4CDwEc5lpGonIuOB8QARERFehGBMjRWzM5PHPtxAyoFcHhrZmYcv6kJtK/5nzlJFzj8vBraragaAiMwGzgOaiEiwe3YRDux2998NtAdS3MtWjYH9pdpLlH7NqdqPo6qTgEkA0dHRWoFjMsZvFBQV8+bibfx9aSLtmtZjxn1DiY5s5nVYxkdV5M+LXcAQEanv9j1cBMQDS4Eb3H3GAXPdx/Pc57jbl6iquu1j3dFSUUAXYBWwGujijq6qg9MJPq8C8RoTMJIycrjh7eW8sSSR6waEM//hCyxRmAqpSJ/FShGZCawFCoF1OH/dfwZMF5GX3bbJ7ksmA/8RkUQgE+fHH1XdJCIzcBJNIfCAqhYBiMiDwAKckVbvquqm8sZrTCBQVaatSualT+OpE1yLf9w6gMvPaeN1WMYPiPPHvf+Ijo7WNWvWeB2GMdVuX04+E2ZtZFFCOud3bs6fb+xrM7HNGRORGFWNPtV2GzNnjB9YnJDGU7M2kp1XyLNX9uSOYZHUsgl2phJZsjDGh+UeLeT3nyXwwcpddG/diA/uGUK31o28Dsv4IUsWxvggVeXL+DRe+jSelANHGD+8I7++tCt1g4O8Ds34KUsWxviY7fsO8/y8TXy9NYMuLRsyffwQhnQM8zos4+csWRjjI3KPFvLW0kTe+WY7dYJr8cwVPRg3LNIm2JlqYcnCmBpOVfk8bi8vfxrPnqw8ruvfjgmju9t62KZaWbIwpgZLTM/h+Xmb+DZxH91bN2LiLf051ybXGQ9YsjCmBsrJL+TNxduY/O126tUJ4oWre3Hr4AiC7ZKT8YglC2NqmEXxafz241jSsvO5cWA4T43uTvOGdb0OywQ4SxbG1BB5BUW88vlm3l++gx5tQvnHrQMZ2KGp12EZA1iyMKZG+CEjh4f+t4741GzuOi+Kp0Z3szkTpkaxZGGMh1SVWWt38+zcOOoG12LyuGgu6tHK67CM+QlLFsZ4JCe/kN99HMecdbsZHNWMiWP7W+E/U2NZsjDGA7EpWTw0bS27MnN57OKuPDiyM0FW+M/UYJYsjKlGqsrkb7fz6hebad6wLtPHD2VQlM2bMDWfJQtjqsn+nHyemLmRJZvTuaRnK167oQ9N6tfxOixjzkiFZviISBMRmSkim0UkQUSGikgzEVkoItvc+6buviIib4hIoohsFJEBpd5nnLv/NhEZV6p9oIjEuq95w12+1Rifoqosik/j8jeW8W3iPl4c04tJtw20RGF8SkWng04EvlDV7kBfIAGYACxW1S7AYvc5wGic9bW7AOOBtwFEpBnwHDAYGAQ8V5Jg3H3uLfW6URWM15hqtX3fYe56fzX3TF1D43q1mfOrYdw+NBL7u8f4mnJfhhKRxsBw4A4AVT0KHBWRMcAId7cpwFfAU8AYYKo667iucM9K2rj7LlTVTPd9FwKjROQrIFRVV7jtU4FrgM/LG7Mx1SX3aCF/X5LIv5dZhVjjHyrSZxEFZADviUhfIAZ4BGilqqnuPnuBkkHj7YDkUq9PcdtO155ykvafEJHxOGcrRERElP+IjKkgVWV+7F5e/iye1Kw8rhvgVohtZENijW+rSLIIBgYAD6nqShGZyLFLTgCoqoqIViTAM6Gqk4BJANHR0VX+ecaczNa0Qzw3dxPfJ+2nZ5tQ3rylP9FWIdb4iYokixQgRVVXus9n4iSLNBFpo6qp7mWmdHf7bqB9qdeHu227OXbZqqT9K7c9/CT7G1OjZOcVMHHRNt5fvoOGdYN56Zre/HxQhM2bMH6l3BdQVXUvkCwi3dymi4B4YB5QMqJpHDDXfTwPuN0dFTUEyHIvVy0ALhWRpm7H9qXAAndbtogMcUdB3V7qvYzxnKoyKyaFkX/+mne/285N0e1Z+vgIbhvSwRKF8TsVnWfxEPCBiNQBkoA7cRLQDBG5G9gJ3OTuOx+4HEgEct19UdVMEXkJWO3u92JJZzfwK+B9oB5Ox7Z1bpsaYc/BIzw1ayPLtu2jX/smvHtHNH3Cm3gdljFVRpzBSf4jOjpa16xZ43UYxk+pKjNjUnjxk3gKi5WnL+/OLwZ3oJadSRgfJyIxqhp9qu02g9uYM5Sencdv5sSyKCGdQZHNeO3GPnQIa+B1WMZUC0sWxpRBVflkYyrPzo3jyNEinrmiB3edF2VnEyagWLIw5jT25+Tzu7lxzI/dS7/2TfjLTX3p1KKh12EZU+0sWRhzCl/E7eW3c2I5lFfIk6O6Mf6CjgTbDGwToCxZGHOCrNwCnpsXx8fr99CrbSj/u7cf3Vo38josYzxlycIYl6ryRdxenpu3iczDR3n04i488LPOVs/JGCxZGANAcmYuz86NY+mWDHq1DeXdO86ld7vGXodlTI1hycIEtKOFxbyzLIk3Fm8juJbw7JU9uX1oB+ubML6jqBB2x8APiyH6LmjUuko+xpKFCVgrk/bz24/jSEzPYXTv1jx3VS9aN7bqsMYHHNgBiYvhhyWw/RvIzwapBW37Q7fRVfKRlixMwMk8fJQ/zk/go5gUwpvW4907ohnZvVXZLzTGK3nZsGOZkxx+WAKZSU574/bQ61roNBKihkP9qqtybMnCBIziYqdUxx8+TyAnr5BfjejEQyO7UK9OkNehGfNT6Zth8yeQuARSVkFxIdRuAFEXwOD7nQQR1hmqadVFSxYmIGxNO8Qzc+JYtSOTcyOb8vtrz6FrKxsOa2oQVdgbCwnzIH4e7NvitLfpC8MedpJD+0EQXNeT8CxZGL9WWFTMv75J4m+LttKwbjB/uqEPNwwIt1IdpmZQdTqn4+c6SeLADqfvocN5MOhe6H4lhLbxOkrAkoXxYz9k5PDrGRtYn3yQK/q04cWrexHW0Ju/yoz5UXER7FrhJIeETyB7N9QKhqgL4fzHnATRoLnXUf6EJQvjd4qLlfeX7+DVLzZTr04Qb97Sn6v6tvU6LBPIiosheQXEzXbOIg6nQ1Bd6HwRjPwddBsF9Zp6HeVpVThZiEgQsAbYrapXikgUMB0IA2KA21T1qIjUBaYCA4H9wM2qusN9j6eBu4Ei4GFVXeC2jwImAkHAv1X1lYrGa/xbcmYuT8zcwIqkTEZ2b8kr151Dy1AbDms8UFwMKath0xyI/xgOpUJwCHS5FHqOga6XQV3f6TerjDOLR4AEINR9/irwuqpOF5F/4iSBt937A6raWUTGuvvdLCI9gbFAL6AtsEhEurrv9RZwCc5636tFZJ6qxldCzMbPqCoz1iTz0qcJAPzp+j7cGB2OVNNIEWMAtw9iLWyaDZs+huwUCKoDnS+B3tf5XIIorULJQkTCgSuA3wP/566VPRL4ubvLFOB5nGQxxn0MMBP4u7v/GGC6quYD20UkERjk7peoqknuZ01397VkYY6Tlp3HhFkbWbolg6Edw/jTDX1o36y+12GZQKEKqRvcBDEHDu6CWrWdS0wX/c6ZJBfi+6VjKnpm8TfgSaAkVYYBB1W10H2eArRzH7cDkgFUtVBEstz92wErSr1n6dckn9A+uILxGj9SsijR7z6OI7+wiOev6sntQyNtpJOpHhlbIW4mxM2C/YlOJ3XHEXDhU9D9ihrfB3G2yp0sRORKIF1VY0RkROWFVK5YxgPjASIiIrwMxVSTrCMF/GZOLJ9tTKV/RBP+cmNfOtqiRKaqHdjpnEHEzoK0WEAg8nwY9hD0uLpKZ1B7rSJnFucBV4vI5UAITp/FRKCJiAS7ZxfhwG53/91AeyBFRIKBxjgd3SXtJUq/5lTtx1HVScAkgOjoaK3AMRkfELMzk4enrSctO48nLuvGfcNtUSJThQ6lOZeX4mY5M6kBws+FUa84pTaqqHBfTVPuZKGqTwNPA7hnFo+r6q0i8hFwA86IqHHAXPcl89zn37vbl6iqisg84H8i8lecDu4uwCpAgC7u6KrdOJ3gJX0hJgAVFStvf5XI64u20bZJCB/dP5T+Ef51qm9qAFXI2OIU6Nv8Cez4FrQYWp0DFz3ndFQ3jfQ6ympXFfMsngKmi8jLwDpgsts+GfiP24GdifPjj6puEpEZOB3XhcADqloEICIPAgtwhs6+q6qbqiBe4wP2ZuXx6IfrWJGUydV92/Lytb0JDantdVjGH6g6hfl2LHMSxPZlzjwIgGadYPgT0Os6aNnd2zg9Jqr+ddUmOjpa16xZ43UYphItik/jiZkbyC8s5oWre3HDQBsSayroYPLxySE7xWlv2Nqp3ho13CnYF0BnECISo6rRp9puM7hNjZVXUMQrnw0I+/AAABf/SURBVG/m/eU76NU2lDdu6U8n68Q25aHqTJCLmw1bv4AD2532+mEQeQFEPeaU26jGKq6+xpKFqZES03N4aNo6ElKzueu8KJ4a3Y26wVZK3JwFVUhd7ySITXMgK9kpsdHpZzD4PidJtOwJtWxwxJmwZGFqlJKZ2M/Pi6denSBbmMicHVVIj3cTxGynL6JWsFPee+Qz0O1yCAkt+33MT1iyMDXGlr2HeG3BZhYlpDOsUxiv39yPVlbXyZyJfduOJYiMzU6Z76jhx6q4+vH8h+piycJ4Likjh78t2sYnG/fQsE4wT4/uzj0XdCTIZmKbUym5xLT5M+eWHg8IdBgGV/wFeoyBhi28jtKvWLIwnknOzOWNxduYvW43dYJqcf+FnRh/QUeaNqjjdWimJioqgJ3fHUsQ2buPLRQ06hWnkmuolaKvKpYsTLXbm5XH35du48PVyYgI44ZG8ssRnWjRyBYmMic4ehgSFznJYesXkJcFwfXcdSCegS6XQYMwr6MMCJYsTLXZl5PPP7/6gf+s2ElRsXLzue15cGRn2jSu53VopibJy4Yt850S30lLoTDPKcrX7QqnQF+nkVDHqgpXN0sWpspl5RYwadkPvPfdDvIKirhuQDiPXNTFyoibY44eds4c4mbDtoVQlA+h4TDwTidBRAyFIPu58pL91zdVpqhY+XB1Mq8t2MyB3AKu6tuWRy/uYhPrjKMgz7nEFDfLSRQFuc4M6ug7off10C7a5kDUIJYsTJVYvSOT5+dtYtOebAZFNeO5q3rSq63vLwBjKqjwKCR95SSILfMhP9uZRd13rFN/qcMwqGWTL2siSxamUu3NyuOPnycwd/0e2jQO4c1b+nNlnzZWyynQ7UuEmPdg/Qdw5ICzclyPq50KrlEX2iUmH2DfkKkUeQVFTP52O28tTaSwWHl4ZGfuH9GJ+nXsn1jAKipwzh5WT4btXzszqbtfAX1vcTqpg230my+x/5NNhagqixLSefmzeHbuz+WyXq145oqe1nkdyLJSIGYKrJ0KOXuhcXtnmGv/26GRlW7xVZYsTLklpufw4qfxfLM1gy4tG/LfuwdzfpfmXodlvFBcDD8sgTWTnc5qVehyCURPdO6tH8LnWbIwZ01V+efXSfzlyy3UqxPEs1f25LahHahtS5sGlqJC2LsRflgMa/8DB3dCgxZw3qMw8A5o2sHrCE0lKneyEJH2wFSgFaDAJFWdKCLNgA+BSGAHcJOqHhCnh3MicDmQC9yhqmvd9xoHPOO+9cuqOsVtHwi8D9QD5gOPqL+t1uRj8gqKeHLmRuZt2MMVfdrwwtW9aN7Qrj0HhOIi2BvrLBq041vYudwZzQTQ4Xy4+DnofhUEW7kWf1SRM4tC4NequlZEGgExIrIQuANYrKqviMgEYALOUqujcdbX7gIMBt4GBrvJ5TkgGifpxIjIPFU94O5zL7ASJ1mMAj6vQMymAlKzjjB+agxxe7J4clQ3fnlhJxvl5M+KiyF9k7OS3I5lTl2mvCxnW1hnZyRT5AUQeT40au1trKbKlTtZqGoqkOo+PiQiCUA7YAwwwt1tCvAVTrIYA0x1zwxWiEgTEWnj7rtQVTMB3IQzSkS+AkJVdYXbPhW4BksWnli76wD3/SeG3PxC3rktmot7WkelX8rPgW0LIH6eM4LpyAGnvWmUU6ivJDlYwb6AUyl9FiISCfTHOQNo5SYSgL04l6nASSTJpV6W4radrj3lJO2mms2KSeHpObG0Dg3hg3sG07VVI69DMpUpLxu2LoD4j50Z1YV50LCVU4spyk0OjcO9jtJ4rMLJQkQaArOAR1U1u/RlCVVVEanyPgYRGQ+MB4iIiKjqjwsYRcXKq19sZtI3SQztGMY/bh1g5cP9xZGDsOVziJ/rdFAXHYVGbZyO6Z5joP1gG8FkjlOhZCEitXESxQeqOtttThORNqqa6l5mSnfbdwPtS7083G3bzbHLViXtX7nt4SfZ/ydUdRIwCSA6Oto6wCtBdl4BD/1vHV9vzWDc0A48c2VPG+3k63IznUly8XPhh6VQXOAU6zv3XidBhJ9rtZjMKVVkNJQAk4EEVf1rqU3zgHHAK+793FLtD4rIdJwO7iw3oSwA/iAiTd39LgWeVtVMEckWkSE4l7duB94sb7zmzCVl5HDP1DXs2p/LH649h58PtrM1n3VwF2yeD5s/dUYvaRE0iYAh90PPa6DdQLBBCuYMVOTM4jzgNiBWRNa7bb/BSRIzRORuYCdwk7ttPs6w2UScobN3ArhJ4SVgtbvfiyWd3cCvODZ09nOsc7vKfbM1gwf/t5bgoFp8cM9gBne0hWV8iqozvHWLmyD2xjrtLXq461FfAW37W4IwZ038bdpCdHS0rlmzxuswfE5RsfKPpYm8vmgrXVs14p3bo61kh68oKoRdy93lRudD1i5AIGKIkxy6XQ5hnbyO0tRwIhKjqtGn2m4zuA0pB3L5vw83sGpHJlf3bcsfrzuHBnXtn0aNdeQg7Fnn3tY6E+SOHICguk6BvgufgK6joWELryM1fsR+EQLcvA17+O2cWFTh9Zv7cm1/GyJZo+TnOCU1dq89lhwyk45tbxoFXUc5Zw+dRkJdW1jKVA1LFgEqJ7+QZ+fGMXvtbgZENOFvN/cnIswuO3kuP8cpxPfDEidB7NsCWuxsC23n9Df0/4Vz36Yf1G/mbbwmYFiyCEBrdx3g0enrSTmQyyMXdeGhkZ0JtmGx3ik4Atu+dJcX/RIKjzirx7Ub6AxpbdvfuVl5b+MhSxYBpKhYeWtpIhMXb6N1aAgz7htKdKT9ZeqJwnxIXAybZjuT447mOBVb+9/qLC8aMdTmPJgaxZJFgEg5kMtjH65n9Y4DjOnXlpeu6U1oSG2vwwosRQWQ9LWTIBI+hfwsqNfUKcjX+3qncqstL2pqKPuX6edUlXkb9vDMx3Gowt9u7sc1/a3EVrU4vA/2rD82cmnX93AkE+qGQvcrnSTRcQQEWdI2NZ8lCz+VV1DE3PW7ee+7HWzee4gBEU2YOLa/zZ2oKrmZkFoqMexZD1ml6mOGdYGul0GPq6DTRVA7xLtYjSkHSxZ+Zs/BI/xnxU6mrdrFwdwCurduxKvXn8P1A8KtE7syZaXAtoVOGe/da51V4ko06+jUWRo03h211AdCGnsXqzGVwJKFH1BVVm3PZMr3O1iwKQ1V5dKerbnjvEgGRzWzBYoqQ+FR5zJS4kLYtggyEpz2Rm2h/bkQfaebGPo6/RDG+BlLFj4sr6CIeev38N7yHSSkZtO4Xm3uuSCK24Z0ILypXW6qsIPJx5LD9q+dEUu1akOHodDv59DlEmjR3eosmYBgycKHZOcVsD3jMEn7cti0O5tZa1M4kFtAt1aNeOW6cxjTrx316tgaBOV2KA1SVsGuFc4iQBmbnfbGEdDnJuh8ibMYUF1b/MkEHksWNczRwmJ2ZeaSlJHD9n2HSco47NzvO8y+nPwf9wuqJVzcoyV3DItiSEe71HTWiosgPR6SV0LyKuf+wA5nW1Ad6DAM+t/mnD0072pnDybgWbLwWFGxsnbXARYlpLF0czo/ZBymqPhYJeDmDesQ1bwBF3VvSVSLBnRs3oCOLRrQvll96gbbWcQZO3IQUtY4Zw7JK53HR3OcbQ1bOSvDnXuPc9+mLwTX9TZeY2oYSxYeOJxfyLJtGSyMT2fplnQyDx+ldpAwOCqMUb1aE9WiAVHNGxLVvAGN69kY/HJRhbRNsG2BM2opeaVTY0lqQave0PcWJzG0H+QsBmRnDsacliWLapKadYTFCeksSkhjeeJ+jhYVExoSzMjuLbm4ZyuGd21hM6orKv+QM0N625dOgji0x2lv0xfO/z+IGu7UW7LKrMactRqfLERkFDARCAL+raqveBzSGSsqVv7z/Q5mrk0hbnc2AB3C6nPb0A5c3KMV0ZFNbV3rilCF/YlOcti6wFk2tLgA6jSCTj9zJsF1vhgatfY6UmN8Xo1OFiISBLwFXAKkAKtFZJ6qxnsbWdmSM3P59QxnQaG+7Zvw1KjuXNyjJZ1bNrTO6IpQhbQ42DQHNn0MmT847S16wJBfQpdLnRXirISGMZWqRicLYBCQqKpJACIyHRgD1NhkoarMWbebZ+duAuAvN/blugHtLEFUhKozcmnTHOe2PxEkyLmsNPRXToJoEuF1lMb4tZqeLNoBpQrskAIM9iiWMh3MPcpv58TxWWwq50Y25a839bNaTBWRnnAsQezb6nROR14AQx90aiw1aO51hMYEjJqeLM6IiIwHxgNERHjzF+a32/bx64/Wsz/nKE+O6sZ9wzsRVMvOJs5YcTHkpDk1lpK+dhJERgIgEHk+DL4PeoyxdaWN8UhNTxa7gfalnoe7bcdR1UnAJIDo6Gg9cXtVyiso4tUvNvPedzvo3LIhk8edS+92VjTuJ4qLIWcvHNzl3na698nOfVYyFB11dxZn8Z/L/ww9rrYV4oypAWp6slgNdBGRKJwkMRb4ubchHRO/J5tHP1zH1rQc7hgWyYTR3QmpbRPlyMuGvbGwdyOkboDUjbB/W6lk4GrQEpq0d6qy9rjS6Xdo0sGZBxHaxpvYjTEnVaOThaoWisiDwAKcobPvquomj8OiqFh5Z1kSf/lyC03q12HKXYO4sGuAXh45vM9JCD8mhg2QmXRse8PWTjLocrGTCJp0cJJC43CoY/05xviKGp0sAFR1PjDf6zj2ZuWxImk/K5L2890P+0jOPMKoXq35w3Xn0KxBHa/Dq177f4BV70DCJ5Cdcqy9SQcnMfT7ObTpB6372CUkY/xEjU8WXknPzuN7NzmsSMpk+77DADQKCWZwVBiPX9qNq/u2DZwhscXFkLQEVk5yJsHVCoZuo2HI/c4M6dbn2DoOxvgxSxaujEP5pZLDfpIySieHZtw6OIIhHcPo0SY0sEY55R+CDdNh5b+cfocGLWHEBBh4p501GBNALFm4XluwmRlrUmhUN5hzo5pxy7lOcujZNsCSQ4nMJOdS07r/Qn42tB0A170DPa+B4AC77GaMsWRR4p4LOvKLIR3o2SY0cNeqLi6G7V85ZxFbF0CtIOh1LQy+H8KjvY7OGOMhSxaurq0qefWzrBTY8jls/cK5lNOgxbFbw5bO7OMGLZzLOg2aO9f7vej/KCqAnd9B/DzY/KkzMa5BC7jwSedSkw1hNcZgyeKYwqNO8bny/mCXFLjbPB+2fOYMIQUI6wyhbZ3LOrtWQO5+4CTzBmsFOz/SrXo5q7R1OA/a9q+aRXgK8+GHpc5opi2fwZEDULu+U6G15xinlIYt/mOMKcWSRYlFz8PaqdAsCpp1/OmtUeufJpKiAqcs9pb5zu3gLkAg/Fy4+HnodgW06Hr8a4qLIDcTDqfD4QxnnkKO+zgnDXavhcUvOvsGhzjv1WGYM6O5/SCo06B8x5ef46wrnTAPtn4JRw9B3VDoOgp6Xg2dLrJ5D8aYU7JkUSJqOBQXOmcAe2OdSzLFhce2B9dzE0eUc8tJd67r5x2EoLrO+gkXPO4MJ23Y8tSfUyvIqW90uhpHh/fDru+d287v4JvXnFXeagU78xc6DHPKcAeHQMERKMyDglzn8Y+33GPtOemw/Rvnef0w6H2tU0Yj6kLrrDbGnBFRrdZSSlUuOjpa16xZU/E3Kip06hVlJsGB7ZC53XmcmeQ8rlPf+au8+xXQaWT5/+I/E3nZztrRO5c7t90xPy2dcaLgEKhdz0lydRtBxwudBBExFILsbwRjzPFEJEZVTzmSxX41TiUo+NhZxImKi537WtU0aiok1OlP6Hyx87wgzzn70WInIZTcgkvuQ6ovNmNMQLBkUR5e/xDXDoH253obgzEmoNifn8YYY8pkycIYY0yZLFkYY4wpkyULY4wxZbJkYYwxpkyWLIwxxpTJkoUxxpgy+d0MbhHJAHaWamoO7PMonKrmr8dmx+V7/PXYAum4OqjqKesQ+V2yOJGIrDndFHZf5q/HZsfle/z12Oy4jrHLUMYYY8pkycIYY0yZAiFZTPI6gCrkr8dmx+V7/PXY7Lhcft9nYYwxpuIC4czCGGNMBVmyMMYYUya/ThYiMkpEtohIoohM8DqeyiIiO0QkVkTWi0glLAvoHRF5V0TSRSSuVFszEVkoItvc+6Zexlgepziu50Vkt/u9rReRy72MsTxEpL2ILBWReBHZJCKPuO0+/Z2d5rj84TsLEZFVIrLBPbYX3PYoEVnp/j5+KCKnXWPZb/ssRCQI2ApcAqQAq4FbVDXe08AqgYjsAKJV1ecnC4nIcCAHmKqqvd22PwGZqvqKm+SbqupTXsZ5tk5xXM8DOar6Zy9jqwgRaQO0UdW1ItIIiAGuAe7Ah7+z0xzXTfj+dyZAA1XNEZHawLfAI8D/AbNVdbqI/BPYoKpvn+p9/PnMYhCQqKpJqnoUmA6M8TgmcwJV/QbIPKF5DDDFfTwF539an3KK4/J5qpqqqmvdx4eABKAdPv6dnea4fJ46ctyntd2bAiOBmW57md+ZPyeLdkByqecp+MmXj/NFfykiMSIy3utgqkArVU11H+8FWnkZTCV7UEQ2upepfOpSzYlEJBLoD6zEj76zE44L/OA7E5EgEVkPpAMLgR+Ag6pa6O5S5u+jPycLf3a+qg4ARgMPuJc8/JI610n95Vrp20AnoB+QCvzF23DKT0QaArOAR1U1u/Q2X/7OTnJcfvGdqWqRqvYDwnGuunQ/2/fw52SxG2hf6nm42+bzVHW3e58OzMH58v1JmnsNueRacrrH8VQKVU1z/6ctBt7BR78397r3LOADVZ3tNvv8d3ay4/KX76yEqh4ElgJDgSYiEuxuKvP30Z+TxWqgi9vjXwcYC8zzOKYKE5EGbgccItIAuBSIO/2rfM48YJz7eBww18NYKk3Jj6nrWnzwe3M7SycDCar611KbfPo7O9Vx+cl31kJEmriP6+EM+knASRo3uLuV+Z357WgoAHeY29+AIOBdVf29xyFVmIh0xDmbAAgG/ufLxyUi04AROCWT04DngI+BGUAETrn5m1TVpzqLT3FcI3AuZyiwA7iv1HV+nyAi5wPLgFig2G3+Dc71fZ/9zk5zXLfg+99ZH5wO7CCcE4QZqvqi+1syHWgGrAN+oar5p3wff04WxhhjKoc/X4YyxhhTSSxZGGOMKZMlC2OMMWWyZGGMMaZMliyM8VMiMkJEhnkdh/EPliyM8V8jAEsWplJYsjB+TUQiRSRBRN5xyzN/6U5MOtm+nUVkkVvKea2IdBLHayISJ05Z+JvdfUeIyNciMldEkkTkFRG51S0FHSsindz93heRf4rIGhHZKiJXuu0hIvKeu+86EfmZ236HiMwWkS/EKff9p1LxXSoi37uxfeSWpigpWf+C2x4rIt3d+kb3A4+JU1r7AhG50T2ODSLyTVX+dzf+J7jsXYzxeV1wytPfKyIzgOuB/55kvw+AV1R1joiE4PwxdR3OpKy+OBPsVpf6oe0L9MCpLpsE/FtVB4mzFsJDwKPufpE4ZSI6AUtFpDPwAE4ZpXNEpDtOYciu7v79cArZ5QNbRORN4AjwDHCxqh4WkadwSky/6L5mn6oOEJFfAY+r6j1u2ekfy2uLSCxwmaruLpnRa8yZsjMLEwi2q+p693EMzo/3cdwSKu1UdQ6Aquapai5wPjDNrQ+UBnwNnOu+bLVb2jofp4rnl2577AmfMUNVi1V1G05S6e6+73/dz9qMM+u5JFksVtUsVc0D4oEOwBCgJ/CdWz10nNteoqRG00mPz/Ud8L6I3Iszm9eYM2ZnFiYQlC5hUASc9DJUBd+3uNTzYo7/f+vEMglllU04Md5gQICFqnpLGa8p2f8nVPV+ERkMXAHEiMhAVd1fRizGAHZmYQzw44I3KSJyDYCI1BWR+jj1gm521wNoAQwHVp3l298oIrXcfoyOwBb3fW91P6srTk2lLad5jxXAee4lrJKCkl1Psz/AIaBRyRMR6aSqK1X1WSCD46syG3NaliyMOeY24GER2QgsB1rjFG3cCGwAlgBPqures3zfXTgJ5nPgfvfy0j+AWm4/wofAHacr4qaqGThLl05z4/uestck+AS4tqSDG3jN7QCPc49vw1kehwlgVkjQmCokIu8Dn6rqzLL2NaYmszMLY4wxZbIzCxNwROQt4LwTmieq6ntexGOML7BkYYwxpkx2GcoYY0yZLFkYY4wpkyULY4wxZbJkYYwxpkyWLIwxxpTJkoUxxpgy/T+wYfh4Qy8uLwAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RKlDyIDsFaxY",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 265
        },
        "outputId": "ce09972f-490a-4848-f81a-d151a47690c1"
      },
      "source": [
        "gmm = GaussianMixture(n_components=5)\n",
        "gmm.fit(df)\n",
        "labels = gmm.predict(df.values)\n",
        "plt.scatter(df.values[:, 0], df.values[:, 6], c=labels, cmap='viridis');"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ic8g1H1gBU-n"
      },
      "source": [
        "scaler = StandardScaler()\n",
        "# Fit on training set only.\n",
        "x = scaler.fit_transform(df)\n",
        "\n",
        "pca = PCA(0.95)\n",
        "X = pca.fit_transform(x)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sFRc0ZG2_r6l"
      },
      "source": [
        "#W2Vec tryout\n",
        "from sklearn.feature_extraction.text import CountVectorizer, TfidfTransformer,TfidfVectorizer\n",
        "from sklearn.naive_bayes import MultinomialNB\n",
        "\n",
        "\n",
        "# count_vect = CountVectorizer()\n",
        "# word_count_vector = count_vect.fit_transform(raw_df['Text'])\n",
        " \n",
        "# settings that you use for count vectorizer will go here\n",
        "#tfidf_vectorizer=TfidfVectorizer(use_idf=True)\n",
        " \n",
        "# just send in all your docs here\n",
        "#tfidf_vectorizer_vectors=tfidf_vectorizer.fit_transform(raw_df['Text'])\n",
        "\n",
        "# tfidf_transformer = TfidfTransformer()\n",
        "# X_raw = tfidf_transformer.fit_transform(X_train_counts)\n",
        "#tfidf_vectorizer_vectors[0].to\n",
        "\n",
        "# for train_index, test_index in kfold.split(X_raw, encoded_raw_Y):\n",
        "#   X_train, y_train = X_raw[train_index], encoded_raw_Y[train_index]\n",
        "#   X_test, y_test = X_raw[test_index], encoded_raw_Y[test_index]\n",
        "#   print(X_raw[train_index])\n",
        "\n",
        "\n",
        "X_train, X_test, y_train, y_test = train_test_split(raw_df['Text'], encoded_raw_Y, random_state = 0)\n",
        "count_vect = CountVectorizer()\n",
        "X_train_counts = count_vect.fit_transform(X_train)\n",
        "tfidf_transformer = TfidfTransformer()\n",
        "X_train_tfidf = tfidf_transformer.fit_transform(X_train_counts)\n",
        "clf = MultinomialNB().fit(X_train_tfidf, y_train)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EEPBnuGac-Jw",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 394
        },
        "outputId": "884bb39e-73ca-43fb-c23b-bb423cc3397d"
      },
      "source": [
        "stopwords = nltk.corpus.stopwords.words('portuguese')\n",
        "#models = [AdaBoostClassifier(), BaggingClassifier(),ExtraTreesClassifier(), GradientBoostingClassifier(),\n",
        "           #RandomForestClassifier(), SGDClassifier(), LinearSVC(), SVC(), BernoulliNB(), MultinomialNB(), MLPClassifier() ]\n",
        "\n",
        "models = [ExtraTreesClassifier(), RandomForestClassifier(), SGDClassifier(), MultinomialNB(), MLPClassifier()]\n",
        "names = ['extratrees', 'randomf', 'sgdclassfier','multinomialNB', 'mlp']\n",
        "\n",
        "results = []\n",
        "accuracies = {}\n",
        "for idx, model in enumerate(models):\n",
        "  for ngram in range(1,5):\n",
        "    clf = make_pipeline(TfidfVectorizer(use_idf=True, stop_words = stopwords, ngram_range=(1, ngram)), model)\n",
        "    accs = []\n",
        "    for i in range(0, 10):\n",
        "      scores = cross_validate(clf, raw_df['Text'], encoded_raw_Y, scoring=['accuracy'], cv=3, return_train_score=True)\n",
        "      #print(np.mean(scores['test_accuracy']))\n",
        "      #print(component, names[idx])\n",
        "      #print(scores['test_accuracy'])\n",
        "      mean = np.mean(scores['test_accuracy'])\n",
        "    accs.append(mean)\n",
        "    results.append((np.mean(accs), ngram, names[idx]))\n",
        "\n",
        "print(results)\n",
        "results"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[(0.21739130434782608, 1, 'extratrees'), (0.24637681159420288, 2, 'extratrees'), (0.2028985507246377, 3, 'extratrees'), (0.2028985507246377, 4, 'extratrees'), (0.2318840579710145, 1, 'randomf'), (0.24637681159420288, 2, 'randomf'), (0.20289855072463767, 3, 'randomf'), (0.14492753623188406, 4, 'randomf'), (0.18840579710144925, 1, 'sgdclassfier'), (0.1884057971014493, 2, 'sgdclassfier'), (0.20289855072463767, 3, 'sgdclassfier'), (0.2028985507246377, 4, 'sgdclassfier'), (0.17391304347826086, 1, 'multinomialNB'), (0.14492753623188406, 2, 'multinomialNB'), (0.14492753623188406, 3, 'multinomialNB'), (0.14492753623188406, 4, 'multinomialNB'), (0.2318840579710145, 1, 'mlp'), (0.17391304347826086, 2, 'mlp'), (0.2028985507246377, 3, 'mlp'), (0.24637681159420288, 4, 'mlp')]\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[(0.21739130434782608, 1, 'extratrees'),\n",
              " (0.24637681159420288, 2, 'extratrees'),\n",
              " (0.2028985507246377, 3, 'extratrees'),\n",
              " (0.2028985507246377, 4, 'extratrees'),\n",
              " (0.2318840579710145, 1, 'randomf'),\n",
              " (0.24637681159420288, 2, 'randomf'),\n",
              " (0.20289855072463767, 3, 'randomf'),\n",
              " (0.14492753623188406, 4, 'randomf'),\n",
              " (0.18840579710144925, 1, 'sgdclassfier'),\n",
              " (0.1884057971014493, 2, 'sgdclassfier'),\n",
              " (0.20289855072463767, 3, 'sgdclassfier'),\n",
              " (0.2028985507246377, 4, 'sgdclassfier'),\n",
              " (0.17391304347826086, 1, 'multinomialNB'),\n",
              " (0.14492753623188406, 2, 'multinomialNB'),\n",
              " (0.14492753623188406, 3, 'multinomialNB'),\n",
              " (0.14492753623188406, 4, 'multinomialNB'),\n",
              " (0.2318840579710145, 1, 'mlp'),\n",
              " (0.17391304347826086, 2, 'mlp'),\n",
              " (0.2028985507246377, 3, 'mlp'),\n",
              " (0.24637681159420288, 4, 'mlp')]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 27
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7HyGlvnLJROc"
      },
      "source": [
        "# Data normalization/scaling\n",
        "scaled_features = PowerTransformer().fit_transform(df.values)\n",
        "df = pd.DataFrame(scaled_features, index=df.index, columns=df.columns)\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CXgZaN4As-nl",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 153
        },
        "outputId": "b5d5dfb1-3cbb-4304-9155-9606a3770021"
      },
      "source": [
        "X, X_test, Y, y_test = train_test_split(df_news.values, encoded_y_news, random_state=7, test_size=0.3)\n",
        "X_test, X_val, y_test, y_val = train_test_split(X_test, y_test, random_state=7, test_size=0.5)\n",
        "\n",
        "model = ExtraTreesClassifier(n_estimators=350, random_state=11)\n",
        "print(X.shape)\n",
        "model.fit(X, Y)\n",
        "\n",
        "print(X_test.shape)\n",
        "y_pred = model.predict(X_test)\n",
        "\n",
        "print(y_pred)\n",
        "print(y_test)\n",
        "score = accuracy_score(y_test, y_pred)\n",
        "print(\"Test score: %.2f%%\" % (score * 100))\n",
        "\n",
        "y_pred = model.predict(X_val)\n",
        "print(y_pred)\n",
        "print(y_val)\n",
        "score = accuracy_score(y_val, y_pred)\n",
        "print(\"Validation score: %.2f%%\" % (score * 100))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(70, 54)\n",
            "(15, 54)\n",
            "[2 6 7 2 7 2 0 2 0 8 2 2 2 4 4]\n",
            "[2 6 6 8 7 0 0 3 0 8 2 8 2 4 4]\n",
            "Test score: 66.67%\n",
            "[3 5 3 1 4 2 1 7 2 2 6 2 8 7 3]\n",
            "[3 5 1 1 4 8 1 7 2 2 6 8 8 2 3]\n",
            "Validation score: 73.33%\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vX53th7hZJSW",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        },
        "outputId": "b6334ccb-4ebf-4117-fcd1-2101da550715"
      },
      "source": [
        "#stopwords = nltk.corpus.stopwords.words('portuguese')\n",
        "models = [ExtraTreesClassifier(), \n",
        "           RandomForestClassifier(), SVC(kernel='poly'), SVC(), MLPClassifier() ]\n",
        "\n",
        "names = ['extratrees','randomf', 'svcpoly', 'svcrbf', 'mlp']\n",
        "\n",
        "#results = []\n",
        "results2 = []\n",
        "accuracies = {}\n",
        "kfold = StratifiedKFold(n_splits=4, shuffle=True, random_state=7)\n",
        "for idx, model in enumerate(models):\n",
        "  for component in range(5, 48):\n",
        "    clf = make_pipeline(PCA(n_components=component), model)\n",
        "    accs = []\n",
        "    for i in range(0, 10):\n",
        "      scores = []\n",
        "      for train_index, test_index in kfold.split(df, encoded_Y):\n",
        "        X_train, y_train = df.iloc[train_index], encoded_Y[train_index]\n",
        "        X_test, y_test = df.iloc[test_index], encoded_Y[test_index]\n",
        "        #X_train_oversampled, y_train_oversampled = SMOTE(k_neighbors=2).fit_sample(X_train, y_train)\n",
        "        model.fit(X_train, y_train)\n",
        "        y_pred = model.predict(X_test)\n",
        "        scores.append(accuracy_score(y_test, y_pred))\n",
        "        #print(np.mean(scores['test_accuracy']))\n",
        "        #print(component, names[idx])\n",
        "        #print(scores['test_accuracy'])\n",
        "      mean = np.mean(scores)\n",
        "      accs.append(mean)\n",
        "    results2.append((np.mean(accs), component, names[idx]))\n",
        "print(results2)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[(0.17663398692810456, 5, 'extratrees'), (0.18815359477124183, 6, 'extratrees'), (0.19133986928104574, 7, 'extratrees'), (0.1793300653594771, 8, 'extratrees'), (0.17058823529411765, 9, 'extratrees'), (0.17908496732026144, 10, 'extratrees'), (0.17181372549019608, 11, 'extratrees'), (0.18341503267973855, 12, 'extratrees'), (0.1853758169934641, 13, 'extratrees'), (0.1883986928104575, 14, 'extratrees'), (0.1901143790849673, 15, 'extratrees'), (0.18235294117647058, 16, 'extratrees'), (0.1935457516339869, 17, 'extratrees'), (0.1849673202614379, 18, 'extratrees'), (0.16911764705882354, 19, 'extratrees'), (0.175, 20, 'extratrees'), (0.19387254901960782, 21, 'extratrees'), (0.17230392156862745, 22, 'extratrees'), (0.1919934640522876, 23, 'extratrees'), (0.18766339869281043, 24, 'extratrees'), (0.16830065359477125, 25, 'extratrees'), (0.1730392156862745, 26, 'extratrees'), (0.18031045751633987, 27, 'extratrees'), (0.18382352941176472, 28, 'extratrees'), (0.17508169934640522, 29, 'extratrees'), (0.18790849673202614, 30, 'extratrees'), (0.1707516339869281, 31, 'extratrees'), (0.20212418300653595, 32, 'extratrees'), (0.18513071895424837, 33, 'extratrees'), (0.18562091503267975, 34, 'extratrees'), (0.1724673202614379, 35, 'extratrees'), (0.1824346405228758, 36, 'extratrees'), (0.19215686274509805, 37, 'extratrees'), (0.1926470588235294, 38, 'extratrees'), (0.16331699346405232, 39, 'extratrees'), (0.19501633986928107, 40, 'extratrees'), (0.1834967320261438, 41, 'extratrees'), (0.19150326797385622, 42, 'extratrees'), (0.1920751633986928, 43, 'extratrees'), (0.1936274509803922, 44, 'extratrees'), (0.18521241830065363, 45, 'extratrees'), (0.1847222222222222, 46, 'extratrees'), (0.1951797385620915, 47, 'extratrees'), (0.19477124183006536, 5, 'randomf'), (0.2051470588235294, 6, 'randomf'), (0.2079248366013072, 7, 'randomf'), (0.20375816993464052, 8, 'randomf'), (0.2053104575163399, 9, 'randomf'), (0.2008169934640523, 10, 'randomf'), (0.2107843137254902, 11, 'randomf'), (0.19060457516339868, 12, 'randomf'), (0.19060457516339868, 13, 'randomf'), (0.1982843137254902, 14, 'randomf'), (0.1946078431372549, 15, 'randomf'), (0.2090686274509804, 16, 'randomf'), (0.20106209150326798, 17, 'randomf'), (0.20767973856209151, 18, 'randomf'), (0.20522875816993466, 19, 'randomf'), (0.19477124183006536, 20, 'randomf'), (0.21086601307189543, 21, 'randomf'), (0.19926470588235295, 22, 'randomf'), (0.19362745098039214, 23, 'randomf'), (0.19207516339869282, 24, 'randomf'), (0.17467320261437908, 25, 'randomf'), (0.19076797385620917, 26, 'randomf'), (0.19501633986928105, 27, 'randomf'), (0.20089869281045752, 28, 'randomf'), (0.20375816993464055, 29, 'randomf'), (0.20359477124183006, 30, 'randomf'), (0.20375816993464052, 31, 'randomf'), (0.19035947712418302, 32, 'randomf'), (0.21495098039215685, 33, 'randomf'), (0.1852124183006536, 34, 'randomf'), (0.2125, 35, 'randomf'), (0.1931372549019608, 36, 'randomf'), (0.20351307189542484, 37, 'randomf'), (0.20359477124183006, 38, 'randomf'), (0.21830065359477122, 39, 'randomf'), (0.20816993464052286, 40, 'randomf'), (0.2125, 41, 'randomf'), (0.20988562091503268, 42, 'randomf'), (0.2019607843137255, 43, 'randomf'), (0.2107843137254902, 44, 'randomf'), (0.1821078431372549, 45, 'randomf'), (0.21666666666666665, 46, 'randomf'), (0.2011437908496732, 47, 'randomf'), (0.13071895424836602, 5, 'svcpoly'), (0.13071895424836602, 6, 'svcpoly'), (0.13071895424836602, 7, 'svcpoly'), (0.13071895424836602, 8, 'svcpoly'), (0.13071895424836602, 9, 'svcpoly'), (0.13071895424836602, 10, 'svcpoly'), (0.13071895424836602, 11, 'svcpoly'), (0.13071895424836602, 12, 'svcpoly'), (0.13071895424836602, 13, 'svcpoly'), (0.13071895424836602, 14, 'svcpoly'), (0.13071895424836602, 15, 'svcpoly'), (0.13071895424836602, 16, 'svcpoly'), (0.13071895424836602, 17, 'svcpoly'), (0.13071895424836602, 18, 'svcpoly'), (0.13071895424836602, 19, 'svcpoly'), (0.13071895424836602, 20, 'svcpoly'), (0.13071895424836602, 21, 'svcpoly'), (0.13071895424836602, 22, 'svcpoly'), (0.13071895424836602, 23, 'svcpoly'), (0.13071895424836602, 24, 'svcpoly'), (0.13071895424836602, 25, 'svcpoly'), (0.13071895424836602, 26, 'svcpoly'), (0.13071895424836602, 27, 'svcpoly'), (0.13071895424836602, 28, 'svcpoly'), (0.13071895424836602, 29, 'svcpoly'), (0.13071895424836602, 30, 'svcpoly'), (0.13071895424836602, 31, 'svcpoly'), (0.13071895424836602, 32, 'svcpoly'), (0.13071895424836602, 33, 'svcpoly'), (0.13071895424836602, 34, 'svcpoly'), (0.13071895424836602, 35, 'svcpoly'), (0.13071895424836602, 36, 'svcpoly'), (0.13071895424836602, 37, 'svcpoly'), (0.13071895424836602, 38, 'svcpoly'), (0.13071895424836602, 39, 'svcpoly'), (0.13071895424836602, 40, 'svcpoly'), (0.13071895424836602, 41, 'svcpoly'), (0.13071895424836602, 42, 'svcpoly'), (0.13071895424836602, 43, 'svcpoly'), (0.13071895424836602, 44, 'svcpoly'), (0.13071895424836602, 45, 'svcpoly'), (0.13071895424836602, 46, 'svcpoly'), (0.13071895424836602, 47, 'svcpoly'), (0.14297385620915032, 5, 'svcrbf'), (0.14297385620915032, 6, 'svcrbf'), (0.14297385620915032, 7, 'svcrbf'), (0.14297385620915032, 8, 'svcrbf'), (0.14297385620915032, 9, 'svcrbf'), (0.14297385620915032, 10, 'svcrbf'), (0.14297385620915032, 11, 'svcrbf'), (0.14297385620915032, 12, 'svcrbf'), (0.14297385620915032, 13, 'svcrbf'), (0.14297385620915032, 14, 'svcrbf'), (0.14297385620915032, 15, 'svcrbf'), (0.14297385620915032, 16, 'svcrbf'), (0.14297385620915032, 17, 'svcrbf'), (0.14297385620915032, 18, 'svcrbf'), (0.14297385620915032, 19, 'svcrbf'), (0.14297385620915032, 20, 'svcrbf'), (0.14297385620915032, 21, 'svcrbf'), (0.14297385620915032, 22, 'svcrbf'), (0.14297385620915032, 23, 'svcrbf'), (0.14297385620915032, 24, 'svcrbf'), (0.14297385620915032, 25, 'svcrbf'), (0.14297385620915032, 26, 'svcrbf'), (0.14297385620915032, 27, 'svcrbf'), (0.14297385620915032, 28, 'svcrbf'), (0.14297385620915032, 29, 'svcrbf'), (0.14297385620915032, 30, 'svcrbf'), (0.14297385620915032, 31, 'svcrbf'), (0.14297385620915032, 32, 'svcrbf'), (0.14297385620915032, 33, 'svcrbf'), (0.14297385620915032, 34, 'svcrbf'), (0.14297385620915032, 35, 'svcrbf'), (0.14297385620915032, 36, 'svcrbf'), (0.14297385620915032, 37, 'svcrbf'), (0.14297385620915032, 38, 'svcrbf'), (0.14297385620915032, 39, 'svcrbf'), (0.14297385620915032, 40, 'svcrbf'), (0.14297385620915032, 41, 'svcrbf'), (0.14297385620915032, 42, 'svcrbf'), (0.14297385620915032, 43, 'svcrbf'), (0.14297385620915032, 44, 'svcrbf'), (0.14297385620915032, 45, 'svcrbf'), (0.14297385620915032, 46, 'svcrbf'), (0.14297385620915032, 47, 'svcrbf'), (0.1929738562091503, 5, 'mlp'), (0.17679738562091502, 6, 'mlp'), (0.19452614379084968, 7, 'mlp'), (0.180718954248366, 8, 'mlp'), (0.17802287581699344, 9, 'mlp'), (0.183578431372549, 10, 'mlp'), (0.1724673202614379, 11, 'mlp'), (0.17540849673202613, 12, 'mlp'), (0.1705065359477124, 13, 'mlp'), (0.19844771241830064, 14, 'mlp'), (0.1793300653594771, 15, 'mlp'), (0.1736111111111111, 16, 'mlp'), (0.19272875816993465, 17, 'mlp'), (0.18562091503267975, 18, 'mlp'), (0.16307189542483658, 19, 'mlp'), (0.1824346405228758, 20, 'mlp'), (0.18276143790849672, 21, 'mlp'), (0.18709150326797386, 22, 'mlp'), (0.18112745098039215, 23, 'mlp'), (0.18112745098039212, 24, 'mlp'), (0.19722222222222224, 25, 'mlp'), (0.17949346405228755, 26, 'mlp'), (0.1826797385620915, 27, 'mlp'), (0.19697712418300653, 28, 'mlp'), (0.18815359477124183, 29, 'mlp'), (0.17802287581699344, 30, 'mlp'), (0.17524509803921567, 31, 'mlp'), (0.18537581699346406, 32, 'mlp'), (0.1652777777777778, 33, 'mlp'), (0.1790849673202614, 34, 'mlp'), (0.19003267973856208, 35, 'mlp'), (0.17230392156862742, 36, 'mlp'), (0.19411764705882356, 37, 'mlp'), (0.1580065359477124, 38, 'mlp'), (0.18251633986928104, 39, 'mlp'), (0.17933006535947715, 40, 'mlp'), (0.1690359477124183, 41, 'mlp'), (0.17352941176470588, 42, 'mlp'), (0.18366013071895423, 43, 'mlp'), (0.172140522875817, 44, 'mlp'), (0.17483660130718953, 45, 'mlp'), (0.17361111111111113, 46, 'mlp'), (0.18120915032679738, 47, 'mlp')]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cHU4EuiXKQV9",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "8f7f4832-b941-493c-b766-8cec8bbeff07"
      },
      "source": [
        "np.mean(accs)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.18120915032679738"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 31
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ez2sJwp__yQm",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "aa935ae0-32fa-45e3-c4a4-aac045a3d372"
      },
      "source": [
        "kfold = StratifiedKFold(n_splits=3, shuffle=True, random_state=7)\n",
        "Y = encoded_Y\n",
        "accuracies = []\n",
        "variances = []\n",
        "X = df\n",
        "\n",
        "models = [ExtraTreesClassifier(), RandomForestClassifier(), MLPClassifier(), SVC(kernel='poly')]\n",
        "for i in range(0,100):\n",
        "  scores = []\n",
        "  for train_index, test_index in kfold.split(X, Y):\n",
        "      #convert integers to dummy variables (i.e. one hot encoded)\n",
        "      #dummy_y = np_utils.to_categorical(Y)\n",
        "      X_train, y_train = X.iloc[train_index], Y[train_index]\n",
        "      X_test, y_test = X.iloc[test_index], Y[test_index]\n",
        "      X_train_oversampled, y_train_oversampled = SMOTE(k_neighbors=1).fit_sample(X_train, y_train)\n",
        "      #rfclf = LinearSVC(multi_class='crammer_singer')\n",
        "      \n",
        "      #rfclf = KNeighborsClassifier(n_neighbors=5)\n",
        "      #rclf = BaggingClassifier(base_estimator=KNeighborsClassifier(n_neighbors=5), n_estimators=100, random_state=0)\n",
        "      #rfclf = ExtraTreesClassifier(max_depth=None, n_estimators=500)\n",
        "      rfclf = LogisticRegression(max_iter=500,solver ='lbfgs')\n",
        "      #rfclf = RandomForestClassifier(n_estimators=500)\n",
        "      rfclf.fit(X_train_oversampled, y_train_oversampled)\n",
        "      y_pred = rfclf.predict(X_test)\n",
        "      #print(confusion_matrix(y_test,y_pred))\n",
        "      #print(classification_report(y_test,y_pred))\n",
        "      #print(accuracy_score(y_test, y_pred))\n",
        "      scores.append(accuracy_score(y_test, y_pred))\n",
        "\n",
        "  accuracies.append(np.mean(scores))\n",
        "print(np.mean(accuracies))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "0.16521739130434782\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0CHmkX29E9dk",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "4d12d954-53af-4dc1-dc10-b95557c49312"
      },
      "source": [
        "from sklearn.model_selection import cross_val_score\n",
        "clf = RandomForestClassifier(n_estimators=200, random_state=0)\n",
        "scores = cross_val_score(clf, X, Y, cv=4)\n",
        "scores"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([0.16666667, 0.11764706, 0.35294118, 0.29411765])"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 33
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "l-LWQrc-6El7",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 323
        },
        "outputId": "44915d48-cc0d-4c29-e23e-7bc7e6d4bae9"
      },
      "source": [
        "# Oversampling \n",
        "from imblearn.over_sampling import SMOTE\n",
        "\n",
        "\n",
        "smt = SMOTE(k_neighbors=1)\n",
        "X_resampled, y_resampled = smt.fit_resample(X_train, y_train)\n",
        "\n",
        "X_resampled, y_resampled"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(array([[-0.55435527, -0.2501882 ,  0.26426935, ...,  1.099827  ,\n",
              "         -0.34048617,  1.09949053],\n",
              "        [ 0.20627463, -0.06161256,  0.58115466, ..., -0.40604434,\n",
              "          0.72685816, -0.94413924],\n",
              "        [ 3.26473613, -2.43843207, -0.27922395, ..., -0.26862725,\n",
              "         -0.96006101,  0.41517329],\n",
              "        ...,\n",
              "        [-0.22854103,  1.1883863 , -0.62574551, ...,  0.45081812,\n",
              "          0.14941078,  1.60373883],\n",
              "        [ 0.45928549, -0.59271705, -0.06906793, ...,  0.55687088,\n",
              "         -0.09401901,  0.15694371],\n",
              "        [ 0.37338248, -0.57500306, -0.10807071, ...,  0.35622545,\n",
              "         -0.00905081,  0.14349321]]),\n",
              " array([ 4,  1,  8,  7,  0,  0,  0,  4, 10,  7,  9,  7,  6,  7,  1,  5,  8,\n",
              "         8,  8,  1,  5,  6,  9,  4,  3,  3,  3,  6,  0, 10,  6,  2,  4,  3,\n",
              "        10,  6,  7,  0, 10,  2,  7,  0, 10,  6,  0,  1,  1,  1,  1,  2,  2,\n",
              "         2,  2,  2,  3,  3,  3,  4,  4,  4,  5,  5,  5,  5,  5,  6,  7,  8,\n",
              "         8,  8,  9,  9,  9,  9,  9, 10, 10]))"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 34
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NNCvBjOW2vYS",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        },
        "outputId": "b1d10915-38b6-48f4-a837-cba15466b43d"
      },
      "source": [
        "# Data split\n",
        "\n",
        "X_train, X_test, y_train, y_test = train_test_split(df, encoded_Y, test_size=0.3, random_state=0)\n",
        "#X_test, X_application, y_test, y_application = train_test_split(X_test, y_test, test_size=0.5, random_state=0)\n",
        "\n",
        "y_train"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([ 8,  9,  7,  7,  0,  2, 10,  6,  0,  6, 10,  4,  9,  4,  0,  5,  7,\n",
              "        7,  7,  4,  1,  8, 10,  1,  1,  0,  7,  8,  6,  6,  2,  7, 10, 10,\n",
              "        0,  1,  9,  7,  8,  0,  5,  3,  6,  5,  4,  0,  3, 10])"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 35
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Od0EJHyl898g",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 527
        },
        "outputId": "ff120bc5-dbff-4a25-9d1b-f18100353f37"
      },
      "source": [
        "#Random Forest baseline + accuracy versus test\n",
        "rfclf = RandomForestClassifier(n_estimators=200, random_state=0)\n",
        "rfclf.fit(X_resampled, y_resampled)\n",
        "y_pred = rfclf.predict(X_test)\n",
        "\n",
        "print(confusion_matrix(y_test,y_pred))\n",
        "print(classification_report(y_test,y_pred))\n",
        "print(accuracy_score(y_test, y_pred))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[[2 0 0 0 0 0 0 0 0 1 0]\n",
            " [0 2 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 1 0 0 0 1 0 0 0 0]\n",
            " [0 0 0 2 0 0 1 0 0 0 0]\n",
            " [0 0 0 0 1 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 1 0 0 0 0 0]\n",
            " [0 0 0 0 0 1 2 0 0 0 0]\n",
            " [0 0 0 0 0 0 1 0 0 0 0]\n",
            " [0 1 0 1 0 0 0 0 1 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 1 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 1]]\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       1.00      0.67      0.80         3\n",
            "           1       0.67      1.00      0.80         2\n",
            "           2       1.00      0.50      0.67         2\n",
            "           3       0.67      0.67      0.67         3\n",
            "           4       1.00      1.00      1.00         1\n",
            "           5       0.50      1.00      0.67         1\n",
            "           6       0.40      0.67      0.50         3\n",
            "           7       0.00      0.00      0.00         1\n",
            "           8       1.00      0.33      0.50         3\n",
            "           9       0.50      1.00      0.67         1\n",
            "          10       1.00      1.00      1.00         1\n",
            "\n",
            "    accuracy                           0.67        21\n",
            "   macro avg       0.70      0.71      0.66        21\n",
            "weighted avg       0.74      0.67      0.65        21\n",
            "\n",
            "0.6666666666666666\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VIq3fFWjEqpV",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 323
        },
        "outputId": "1f1141e0-9711-496f-9c52-83aff5feca35"
      },
      "source": [
        "# Random Forest hyperparam discovery\n",
        "\n",
        "# Number of trees in random forest\n",
        "n_estimators = [int(x) for x in np.linspace(start = 10, stop = 800, num = 10)]\n",
        "# Number of features to consider at every split\n",
        "#max_features = ['auto', 'sqrt']\n",
        "loss = ['deviance', 'exponential']\n",
        "learning_rate = [0.1, 0.2, 0.3, 0.4,0.5]\n",
        "# Maximum number of levels in tree\n",
        "max_depth = [int(x) for x in np.linspace(10, 110, num = 11)]\n",
        "max_depth.append(None)\n",
        "# Minimum number of samples required to split a node\n",
        "min_samples_split = [0, 1, 2, 5, 10]\n",
        "# Minimum number of samples required at each leaf node\n",
        "min_samples_leaf = [0, 1, 2, 4]\n",
        "# Method of selecting samples for training each tree\n",
        "bootstrap = [True, False]\n",
        "# Create the random grid\n",
        "random_grid = {'n_estimators': n_estimators,\n",
        "               #'loss': loss,\n",
        "               'max_depth': max_depth,\n",
        "               'min_samples_split': min_samples_split,\n",
        "               'min_samples_leaf': min_samples_leaf}\n",
        "               #'learning_rate': learning_rate}\n",
        "               #'bootstrap': bootstrap}\n",
        "\n",
        "kfold = StratifiedKFold(n_splits=3, shuffle=True, random_state=7)\n",
        "Y = encoded_Y\n",
        "scores = []\n",
        "model = []\n",
        "best_params = []\n",
        "X = df\n",
        "for train_index, test_index in kfold.split(X, Y):\n",
        "    #convert integers to dummy variables (i.e. one hot encoded)\n",
        "    #dummy_y = np_utils.to_categorical(Y)\n",
        "    X_train, y_train = X.iloc[train_index], Y[train_index]\n",
        "    X_test, y_test = X.iloc[test_index], Y[test_index]\n",
        "    X_train_oversampled, y_train_oversampled = SMOTE(k_neighbors=1).fit_sample(X_train, y_train)\n",
        "    hrfclf = ExtraTreesClassifier()\n",
        "    rf_random = RandomizedSearchCV(estimator = hrfclf, param_distributions = random_grid, n_iter = 100, cv = 3, verbose=2, random_state=42, n_jobs = -1)\n",
        "    rf_random.fit(X_train_oversampled, y_train_oversampled)\n",
        "    model.append(rf_random.best_estimator_)\n",
        "    best_params.append(rf_random.best_params_)\n",
        "    y_pred = rf_random.best_estimator_.predict(X_test)\n",
        "    #print(confusion_matrix(y_test,y_pred))\n",
        "    #print(classification_report(y_test,y_pred))\n",
        "    print(accuracy_score(y_test, y_pred))\n",
        "    scores.append(accuracy_score(y_test, y_pred))\n",
        "\n",
        "\n",
        "# Fit the random search model\n",
        "#rf_random.fit(X_resampled, y_resampled)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Fitting 3 folds for each of 100 candidates, totalling 300 fits\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 2 concurrent workers.\n",
            "[Parallel(n_jobs=-1)]: Done  37 tasks      | elapsed:   12.4s\n",
            "[Parallel(n_jobs=-1)]: Done 158 tasks      | elapsed:   37.2s\n",
            "[Parallel(n_jobs=-1)]: Done 300 out of 300 | elapsed:  1.1min finished\n",
            "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 2 concurrent workers.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "0.043478260869565216\n",
            "Fitting 3 folds for each of 100 candidates, totalling 300 fits\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[Parallel(n_jobs=-1)]: Done  37 tasks      | elapsed:   11.5s\n",
            "[Parallel(n_jobs=-1)]: Done 158 tasks      | elapsed:   36.5s\n",
            "[Parallel(n_jobs=-1)]: Done 300 out of 300 | elapsed:  1.1min finished\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "0.21739130434782608\n",
            "Fitting 3 folds for each of 100 candidates, totalling 300 fits\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 2 concurrent workers.\n",
            "[Parallel(n_jobs=-1)]: Done  37 tasks      | elapsed:   11.4s\n",
            "[Parallel(n_jobs=-1)]: Done 158 tasks      | elapsed:   36.3s\n",
            "[Parallel(n_jobs=-1)]: Done 300 out of 300 | elapsed:  1.1min finished\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "0.2608695652173913\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VeWYLSyoHe0B",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 221
        },
        "outputId": "953a6f76-dadf-4a7c-9619-029e2a90883a"
      },
      "source": [
        "best_params"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[{'max_depth': None,\n",
              "  'min_samples_leaf': 2,\n",
              "  'min_samples_split': 2,\n",
              "  'n_estimators': 97},\n",
              " {'max_depth': None,\n",
              "  'min_samples_leaf': 1,\n",
              "  'min_samples_split': 2,\n",
              "  'n_estimators': 536},\n",
              " {'max_depth': 30,\n",
              "  'min_samples_leaf': 1,\n",
              "  'min_samples_split': 5,\n",
              "  'n_estimators': 448}]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 38
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "tH3TNQD0Hvt0",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "0c52d687-fed8-4dd6-9942-57547690f9c2"
      },
      "source": [
        "y_pred = rf_random.best_estimator_.predict(X_test)\n",
        "print(accuracy_score(y_test, y_pred))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "0.2608695652173913\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "l-2p_e_F_Pnd",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "44a4f9e6-278b-4193-c094-3ad5007171a9"
      },
      "source": [
        "# RFC w/ tunned HP\n",
        "rfclf = RandomForestClassifier(bootstrap='True', max_depth=20, max_features='sqrt', n_estimators=900, random_state=0)\n",
        "rfclf.fit(X_resampled, y_resampled)\n",
        "pred= rfclf.predict(X_test)\n",
        "accuracy_score(pred, y_test)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.2608695652173913"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 40
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0HB3cf2La26x"
      },
      "source": [
        "glove_100_link = 'https://drive.google.com/open?id=1hW4DhKFWJJPNIEI3GelC1WLzEqU-lrub'\n",
        "\n",
        "glove_50_link = 'https://drive.google.com/open?id=10MW2F53DYYxizSXLbvfsphs77Fu1U9Gz'\n",
        "\n",
        "skip_text_link = 'https://drive.google.com/open?id=1WXpYIhDI8Hhne3cOZUQshScjwsWO2wRR'\n",
        "\n",
        "cbow_link = 'https://drive.google.com/open?id=1o6bO6QDAOFd_bASWuZHfbBlk8OBNA0uQ'\n",
        "\n",
        "def load_file(link, filename):\n",
        "  # Download gloVe word embedding for portuguese-BR (100-dimensions)\n",
        "  auth.authenticate_user()\n",
        "  gauth = GoogleAuth()\n",
        "  gauth.credentials = GoogleCredentials.get_application_default()\n",
        "  drive = GoogleDrive(gauth)\n",
        "  fluff, id = link.split('=')\n",
        "  downloaded = drive.CreateFile({'id':id})\n",
        "  downloaded.GetContentFile(filename)\n",
        "\n",
        "def build_embedding_dict(filename):\n",
        "  embd_dict = {}\n",
        "  with open(filename, 'r', encoding='utf-8') as f:\n",
        "    for line in f:\n",
        "      word, coefs = line.split(maxsplit=1)\n",
        "      coefs = np.fromstring(coefs, 'f', sep=' ')\n",
        "      embd_dict[word] = coefs\n",
        "  return embd_dict\n",
        "  "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3DK4tokfAYo3"
      },
      "source": [
        "# Download and load embeddings into dictionaries (eg: {'oi' : n1, n2, ..., n100})\n",
        "\n",
        "load_file(glove_100_link, 'glove_100.txt')\n",
        "load_file(glove_50_link, 'glove_50.txt')\n",
        "load_file(cbow_link, 'cbow_50.txt')\n",
        "load_file(skip_text_link, 'skipgram_50.txt')\n",
        "\n",
        "word_embedding_dict_50 = build_embedding_dict('glove_50.txt')\n",
        "word_embedding_dict_100 = build_embedding_dict('glove_100.txt')\n",
        "word_embedding_cbow_50 = build_embedding_dict('cbow_50.txt')\n",
        "word_embedding_skipgram_50 = build_embedding_dict('skipgram_50.txt')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "abWAExN5hmuy"
      },
      "source": [
        "corpus = raw_df['Text']\n",
        "\n",
        "# Create a Keras tokenizer and convert words to integer values\n",
        "tokenizer = Tokenizer(num_words=20000)\n",
        "tokenizer.fit_on_texts(corpus)\n",
        "\n",
        "# Convert strings to their corresponding integer value created by the Tokenizer\n",
        "embedded_sentences = tokenizer.texts_to_sequences(corpus)\n",
        "print(embedded_sentences)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6yU349i7__41"
      },
      "source": [
        "import pandas as pd\n",
        "\n",
        "rows = pd.Series(['Regressão Logística', 'Naive Bayes MN', 'Naive Bayer Gausiano', 'SVM Linear', 'SVM RBF', 'SVM Polinomial', 'Random Forest',\n",
        "              'Árvores Extra','Árvore com gradiente descendente', 'MLP'])\n",
        "\n",
        "cols = pd.Series(['acc', 'roc auc'])\n",
        "\n",
        "df = pd.DataFrame(index=rows, columns=cols)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 359
        },
        "id": "F3lUSFBdNNHe",
        "outputId": "80802cd4-a87a-4491-b1fb-6a64513c364d"
      },
      "source": [
        "df.loc['Regressão Logística']['acc'] = 10"
      ],
      "execution_count": null,
      "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>acc</th>\n",
              "      <th>roc auc</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>Regressão Logística</th>\n",
              "      <td>10</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Naive Bayes MN</th>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Naive Bayer Gausiano</th>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>SVM Linear</th>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>SVM RBF</th>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>SVM Polinomial</th>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Random Forest</th>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Árvores Extra</th>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Árvore com gradiente descendente</th>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>MLP</th>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                                  acc roc auc\n",
              "Regressão Logística                10     NaN\n",
              "Naive Bayes MN                    NaN     NaN\n",
              "Naive Bayer Gausiano              NaN     NaN\n",
              "SVM Linear                        NaN     NaN\n",
              "SVM RBF                           NaN     NaN\n",
              "SVM Polinomial                    NaN     NaN\n",
              "Random Forest                     NaN     NaN\n",
              "Árvores Extra                     NaN     NaN\n",
              "Árvore com gradiente descendente  NaN     NaN\n",
              "MLP                               NaN     NaN"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 9
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BBcappPQjXYf",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "d13a0af3-add7-42ed-f769-ad6874951f2d"
      },
      "source": [
        "max_words_text = max(corpus, key=lambda sentence: len(nltk.word_tokenize(sentence, language='portuguese')))\n",
        "print(max_words_text[:200])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "biologia tendências perspectivas atuais abordagem sala aula texto disponibilizado sala aula virtual, histórico ciências, deve ter percebido modificações ocorridas estreita relação transformações polít\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "R_jRhtYJjfkf",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "9b90254c-48b0-4e68-f6c8-e3e270508d83"
      },
      "source": [
        "max_sentence_len = len(nltk.word_tokenize(max_words_text, language='portuguese'))\n",
        "print(max_sentence_len)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "3913\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XDmfjQ2tjg4T",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "2cc8d610-1d7f-499a-b17d-cb2f571954fc"
      },
      "source": [
        "# Make all the sentences have the same length by padding with 0s at the end\n",
        "padded_sentences = pad_sequences(embedded_sentences, max_sentence_len, padding='post')\n",
        "\n",
        "padded_sentences.shape"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(69, 3913)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 47
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rfw8qPMTj3_a"
      },
      "source": [
        "vocab_len = len(tokenizer.word_index) + 1\n",
        "\n",
        "# Create an embedding matrix, in other words, use the weights from gloVe vector against all our corpus\n",
        "# In the end we will have a matrix of all our vocabulary x N dimensions\n",
        "def build_embd_matrix(size, embd_dic):\n",
        "  # initialize with all zeros\n",
        "  embedding_matrix = np.zeros((vocab_len, size))\n",
        "  for word, index in tokenizer.word_index.items():\n",
        "    embedding_vector = embd_dic.get(word)\n",
        "    # if the word is in gloVe, we will use its weigths \n",
        "    if embedding_vector is not None:\n",
        "      embedding_matrix[index] = embedding_vector[:size]\n",
        "  return embedding_matrix\n",
        "\n",
        "\n",
        "embedding_matrix_50 = build_embd_matrix(50, word_embedding_dict_50)\n",
        "\n",
        "embedding_matrix_100 = build_embd_matrix(100, word_embedding_dict_100)\n",
        "\n",
        "embedding_matrix_cbow_50 = build_embd_matrix(50, word_embedding_cbow_50)\n",
        "\n",
        "embedding_matrix_skipgram_50 = build_embd_matrix(50, word_embedding_skipgram_50)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EBbXSjHcJf3N",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 357
        },
        "outputId": "ab080bb5-3356-4664-9e8b-2e146a9d8dfd"
      },
      "source": [
        "# TODO: Update this model to user TensorFlow 2.x\n",
        "model = Sequential()\n",
        "model.add(Embedding(vocab_len, 50,weights=[embedding_matrix_50], input_length=max_sentence_len))\n",
        "model.add(Dropout(0.5))\n",
        "model.add(Conv1D(64, 5, activation='relu'))\n",
        "\n",
        "model.add(MaxPool1D(2, strides=2))\n",
        "model.add(Flatten())\n",
        "model.add(Dense(11, activation='softmax'))\n",
        "model.summary()\n",
        "model.compile(loss='categorical_crossentropy', optimizer='adam',metrics=['acc'])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_1\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "embedding (Embedding)        (None, 3913, 50)          250200    \n",
            "_________________________________________________________________\n",
            "dropout (Dropout)            (None, 3913, 50)          0         \n",
            "_________________________________________________________________\n",
            "conv1d (Conv1D)              (None, 3909, 64)          16064     \n",
            "_________________________________________________________________\n",
            "max_pooling1d (MaxPooling1D) (None, 1954, 64)          0         \n",
            "_________________________________________________________________\n",
            "flatten (Flatten)            (None, 125056)            0         \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, 11)                1375627   \n",
            "=================================================================\n",
            "Total params: 1,641,891\n",
            "Trainable params: 1,641,891\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "L4CmlnXmKe22",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "6ac20d2f-fd03-4576-a7a7-e1411f003771"
      },
      "source": [
        "X = padded_sentences\n",
        "Y = np_utils.to_categorical(encoded_Y)\n",
        "model.fit(X, Y, epochs=50, verbose=1)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/50\n",
            "3/3 [==============================] - 1s 187ms/step - loss: 5.0218 - acc: 0.0870\n",
            "Epoch 2/50\n",
            "3/3 [==============================] - 1s 182ms/step - loss: 2.5624 - acc: 0.2609\n",
            "Epoch 3/50\n",
            "3/3 [==============================] - 1s 181ms/step - loss: 1.7085 - acc: 0.4638\n",
            "Epoch 4/50\n",
            "3/3 [==============================] - 1s 181ms/step - loss: 1.0356 - acc: 0.6812\n",
            "Epoch 5/50\n",
            "3/3 [==============================] - 1s 177ms/step - loss: 0.8874 - acc: 0.7681\n",
            "Epoch 6/50\n",
            "3/3 [==============================] - 1s 178ms/step - loss: 0.8151 - acc: 0.7681\n",
            "Epoch 7/50\n",
            "3/3 [==============================] - 1s 180ms/step - loss: 0.7911 - acc: 0.8406\n",
            "Epoch 8/50\n",
            "3/3 [==============================] - 1s 178ms/step - loss: 0.4871 - acc: 0.8261\n",
            "Epoch 9/50\n",
            "3/3 [==============================] - 1s 178ms/step - loss: 0.6800 - acc: 0.8696\n",
            "Epoch 10/50\n",
            "3/3 [==============================] - 1s 179ms/step - loss: 0.3977 - acc: 0.9275\n",
            "Epoch 11/50\n",
            "3/3 [==============================] - 1s 182ms/step - loss: 0.3537 - acc: 0.9130\n",
            "Epoch 12/50\n",
            "3/3 [==============================] - 1s 178ms/step - loss: 0.3576 - acc: 0.9420\n",
            "Epoch 13/50\n",
            "3/3 [==============================] - 1s 179ms/step - loss: 0.2597 - acc: 0.9565\n",
            "Epoch 14/50\n",
            "3/3 [==============================] - 1s 177ms/step - loss: 0.2264 - acc: 0.9565\n",
            "Epoch 15/50\n",
            "3/3 [==============================] - 1s 179ms/step - loss: 0.1925 - acc: 0.9710\n",
            "Epoch 16/50\n",
            "3/3 [==============================] - 1s 177ms/step - loss: 0.1363 - acc: 0.9855\n",
            "Epoch 17/50\n",
            "3/3 [==============================] - 1s 177ms/step - loss: 0.1330 - acc: 0.9855\n",
            "Epoch 18/50\n",
            "3/3 [==============================] - 1s 178ms/step - loss: 0.0977 - acc: 0.9855\n",
            "Epoch 19/50\n",
            "3/3 [==============================] - 1s 187ms/step - loss: 0.1264 - acc: 0.9855\n",
            "Epoch 20/50\n",
            "3/3 [==============================] - 1s 179ms/step - loss: 0.1349 - acc: 0.9855\n",
            "Epoch 21/50\n",
            "3/3 [==============================] - 1s 177ms/step - loss: 0.0947 - acc: 0.9855\n",
            "Epoch 22/50\n",
            "3/3 [==============================] - 1s 175ms/step - loss: 0.0810 - acc: 0.9855\n",
            "Epoch 23/50\n",
            "3/3 [==============================] - 1s 177ms/step - loss: 0.0739 - acc: 0.9855\n",
            "Epoch 24/50\n",
            "3/3 [==============================] - 1s 177ms/step - loss: 0.0656 - acc: 0.9855\n",
            "Epoch 25/50\n",
            "3/3 [==============================] - 1s 179ms/step - loss: 0.0627 - acc: 0.9855\n",
            "Epoch 26/50\n",
            "3/3 [==============================] - 1s 180ms/step - loss: 0.0728 - acc: 0.9855\n",
            "Epoch 27/50\n",
            "3/3 [==============================] - 1s 177ms/step - loss: 0.0546 - acc: 0.9855\n",
            "Epoch 28/50\n",
            "3/3 [==============================] - 1s 180ms/step - loss: 0.0430 - acc: 0.9855\n",
            "Epoch 29/50\n",
            "3/3 [==============================] - 1s 188ms/step - loss: 0.0399 - acc: 0.9855\n",
            "Epoch 30/50\n",
            "3/3 [==============================] - 1s 179ms/step - loss: 0.0352 - acc: 0.9710\n",
            "Epoch 31/50\n",
            "3/3 [==============================] - 1s 181ms/step - loss: 0.0320 - acc: 1.0000\n",
            "Epoch 32/50\n",
            "3/3 [==============================] - 1s 180ms/step - loss: 0.0319 - acc: 0.9855\n",
            "Epoch 33/50\n",
            "3/3 [==============================] - 1s 182ms/step - loss: 0.0269 - acc: 0.9855\n",
            "Epoch 34/50\n",
            "3/3 [==============================] - 1s 181ms/step - loss: 0.0755 - acc: 0.9855\n",
            "Epoch 35/50\n",
            "3/3 [==============================] - 1s 185ms/step - loss: 0.0474 - acc: 0.9855\n",
            "Epoch 36/50\n",
            "3/3 [==============================] - 1s 180ms/step - loss: 0.0375 - acc: 0.9855\n",
            "Epoch 37/50\n",
            "3/3 [==============================] - 1s 182ms/step - loss: 0.0563 - acc: 0.9855\n",
            "Epoch 38/50\n",
            "3/3 [==============================] - 1s 180ms/step - loss: 0.0582 - acc: 0.9855\n",
            "Epoch 39/50\n",
            "3/3 [==============================] - 1s 180ms/step - loss: 0.0435 - acc: 0.9855\n",
            "Epoch 40/50\n",
            "3/3 [==============================] - 1s 180ms/step - loss: 0.0400 - acc: 0.9855\n",
            "Epoch 41/50\n",
            "3/3 [==============================] - 1s 180ms/step - loss: 0.0295 - acc: 0.9855\n",
            "Epoch 42/50\n",
            "3/3 [==============================] - 1s 181ms/step - loss: 0.0309 - acc: 1.0000\n",
            "Epoch 43/50\n",
            "3/3 [==============================] - 1s 179ms/step - loss: 0.0208 - acc: 1.0000\n",
            "Epoch 44/50\n",
            "3/3 [==============================] - 1s 182ms/step - loss: 0.0260 - acc: 0.9855\n",
            "Epoch 45/50\n",
            "3/3 [==============================] - 1s 180ms/step - loss: 0.0961 - acc: 0.9855\n",
            "Epoch 46/50\n",
            "3/3 [==============================] - 1s 180ms/step - loss: 0.0286 - acc: 0.9855\n",
            "Epoch 47/50\n",
            "3/3 [==============================] - 1s 181ms/step - loss: 0.0389 - acc: 0.9855\n",
            "Epoch 48/50\n",
            "3/3 [==============================] - 1s 179ms/step - loss: 0.0386 - acc: 0.9855\n",
            "Epoch 49/50\n",
            "3/3 [==============================] - 1s 177ms/step - loss: 0.0464 - acc: 0.9855\n",
            "Epoch 50/50\n",
            "3/3 [==============================] - 1s 178ms/step - loss: 0.0247 - acc: 0.9855\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tensorflow.python.keras.callbacks.History at 0x7f67d208b6a0>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 53
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "34ozDaDvdWKa"
      },
      "source": [
        "def build_model(embd_matrix, size=50):\n",
        "  model = Sequential()\n",
        "  embedding_layer = Embedding(vocab_len, 50, input_length=max_sentence_len, \n",
        "                              embeddings_initializer=Constant(embedding_matrix_50), trainable=False)\n",
        "  model.add(embedding_layer)\n",
        "  model.add(Dropout(0.3))\n",
        "  model.add(Flatten())\n",
        "  number_of_classes = len(np_utils.to_categorical(encoded_raw_Y)[0])\n",
        "  model.add(Dense(number_of_classes, activation='softmax'))\n",
        "\n",
        "  model.compile(optimizer='rmsprop',\n",
        "                    loss='categorical_crossentropy', metrics=['accuracy'])\n",
        "  return model"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "71gTUZklld6b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 544
        },
        "outputId": "13131bd6-a14f-4f1e-913d-98cbc9cd37b2"
      },
      "source": [
        "cv_scores = []\n",
        "kfold = StratifiedKFold(n_splits=3, shuffle=True, random_state=7)\n",
        "models = []\n",
        "\n",
        "X = padded_sentences\n",
        "\n",
        "X, X_val, Y_train, y_val = train_test_split(padded_sentences, encoded_raw_Y, stratify=encoded_raw_Y, random_state=11, test_size=0.20)\n",
        "\n",
        "accs = []\n",
        "for i in range(0,10):\n",
        "  # stratified k-fold validation\n",
        "  for train_index, test_index in kfold.split(X, Y_train):\n",
        "    model = build_model(embd_matrix=embedding_matrix_100)\n",
        "    \n",
        "    # Convert integers to vectors (e.g One hot encoding)\n",
        "    Y = np_utils.to_categorical(Y_train)\n",
        "\n",
        "    #print(\"TRAIN:\", train_index, \"TEST:\", test_index)\n",
        "    X_train, X_test = X[train_index], X[test_index]\n",
        "    y_train, y_test = Y[train_index], Y[test_index]\n",
        "\n",
        "    # train the model\n",
        "    model.fit(X_train, y_train, epochs=50, verbose=0)\n",
        "\n",
        "    # Evaluate the model\n",
        "    scores = model.evaluate(X_test, y_test)\n",
        "    #print(\"%s: %.2f%%\" % (model.metrics_names[1], scores[1] * 100))\n",
        "    cv_scores.append(scores[1] * 100)\n",
        "  accs.append(np.mean(cv_scores))\n",
        "\n",
        "# Brief summary of the scores\n",
        "print(\"%.2f%% (+/- %.2f%%)\" % (np.mean(accs), np.std(accs)))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "1/1 [==============================] - 0s 2ms/step - loss: 3.9614 - accuracy: 0.2105\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 2.7498 - accuracy: 0.3333\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 4.3058 - accuracy: 0.3889\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 3.9903 - accuracy: 0.2105\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 3.3617 - accuracy: 0.2778\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 4.5151 - accuracy: 0.2778\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 3.6268 - accuracy: 0.2105\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 2.9173 - accuracy: 0.2778\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 4.7232 - accuracy: 0.2778\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 3.6035 - accuracy: 0.1579\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 3.0039 - accuracy: 0.1667\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 4.7960 - accuracy: 0.1667\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 3.6179 - accuracy: 0.2105\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 3.3036 - accuracy: 0.2222\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 4.0724 - accuracy: 0.1667\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 3.6237 - accuracy: 0.2105\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 3.2248 - accuracy: 0.1667\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 4.3326 - accuracy: 0.2778\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 3.4954 - accuracy: 0.2105\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 3.4534 - accuracy: 0.2222\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 4.2424 - accuracy: 0.2778\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 4.0016 - accuracy: 0.2105\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 3.2591 - accuracy: 0.2222\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 4.3005 - accuracy: 0.1667\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 4.0055 - accuracy: 0.1579\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 3.4227 - accuracy: 0.1667\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 5.4700 - accuracy: 0.2222\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 3.5081 - accuracy: 0.1579\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 3.7841 - accuracy: 0.2222\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 4.5791 - accuracy: 0.2222\n",
            "24.97% (+/- 2.81%)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "22bWRKETwqEc",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "3f95e9ef-75fa-469f-bc0d-ed592fb6f2c8"
      },
      "source": [
        "y_val.shape\n",
        "y_val_dummy = np_utils.to_categorical(y_val)\n",
        "X_val.shape"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(14, 3913)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 58
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "F5oTvDXl0lwZ",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        },
        "outputId": "0bcc01cd-505a-4e58-ea25-7c9eebeae97d"
      },
      "source": [
        "scores = model.evaluate(X_val, y_val_dummy)\n",
        "print(\"Test score: %.2f%%\" % (scores[1] * 100))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "1/1 [==============================] - 0s 1ms/step - loss: 7.8574 - accuracy: 0.0714\n",
            "Test score: 7.14%\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "82-C9horrGZj"
      },
      "source": [
        "#scores = model.evaluate(X_val, y_val_dummy)\n",
        "\n",
        "X, X_val, Y_train, y_val = train_test_split(padded_sentences, encoded_raw_Y, random_state=11, stratify=encoded_raw_Y, test_size=0.20)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "o-jx9oR2kbNu"
      },
      "source": [
        "def test_model(iter, embedding, name, epochs):\n",
        "  test_scores = []\n",
        "  val_scores = []\n",
        "  for i in range(iter):\n",
        "    Y = np_utils.to_categorical(encoded_raw_Y)\n",
        "\n",
        "    X, X_test, Y, y_test = train_test_split(padded_sentences, Y, stratify=Y, random_state=11, test_size=0.2)\n",
        "    X_test, X_val, y_test, y_val = train_test_split(X_test, y_test, random_state=11, test_size=0.5)\n",
        "\n",
        "    model = build_model(embedding, 100)\n",
        "\n",
        "    model.fit(X, Y, epochs=epochs, verbose=0)\n",
        "    scores = model.evaluate(X_test, y_test)\n",
        "\n",
        "    #print(\"Test score \" + name + \": %.2f%%\" % (scores[1] * 100))\n",
        "    test_scores.append(scores[1] * 100)\n",
        "\n",
        "    scores = model.evaluate(X_val, y_val)\n",
        "    #print(\"Validation score \" + name + \": %.2f%%\" % (scores[1] * 100))\n",
        "    val_scores.append(scores[1] * 100)\n",
        "  print(\"Mean test scores:\" + \": %.2f%%\" % (np.mean(test_scores)))\n",
        "  print(\"Mean val scores:\" + \": %.2f%%\" % (np.mean(val_scores)))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nwC-e46eh4vf",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "60e91a34-d4f3-4e41-d394-7b6a68025732"
      },
      "source": [
        "# Run and print models output\n",
        "test_model(10,embedding_matrix_50, 'Glove 50D', 150)\n",
        "test_model(10,embedding_matrix_100,'Glove 100D', 150 )\n",
        "test_model(10,embedding_matrix_cbow_50, 'CBOW 50D', 150)\n",
        "test_model(10,embedding_matrix_skipgram_50, 'SkipGram 50D', 150)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "1/1 [==============================] - 0s 1ms/step - loss: 6.2562 - accuracy: 0.0000e+00\n",
            "Test score Glove 100D: 0.00%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 2.7050 - accuracy: 0.1429\n",
            "Validation score Glove 100D: 14.29%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 6.2642 - accuracy: 0.0000e+00\n",
            "Test score Glove 100D: 0.00%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 2.8893 - accuracy: 0.2857\n",
            "Validation score Glove 100D: 28.57%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 5.5345 - accuracy: 0.1429\n",
            "Test score Glove 100D: 14.29%\n",
            "1/1 [==============================] - 0s 949us/step - loss: 2.4820 - accuracy: 0.1429\n",
            "Validation score Glove 100D: 14.29%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 5.9260 - accuracy: 0.1429\n",
            "Test score Glove 100D: 14.29%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 2.6927 - accuracy: 0.1429\n",
            "Validation score Glove 100D: 14.29%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 6.5702 - accuracy: 0.0000e+00\n",
            "Test score Glove 100D: 0.00%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 2.5701 - accuracy: 0.1429\n",
            "Validation score Glove 100D: 14.29%\n",
            "Mean test scores:: 5.71%\n",
            "Mean val scores:: 20.00%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 6.3067 - accuracy: 0.0000e+00\n",
            "Test score CBOW 50D: 0.00%\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 2.5512 - accuracy: 0.1429\n",
            "Validation score CBOW 50D: 14.29%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 6.1061 - accuracy: 0.1429\n",
            "Test score CBOW 50D: 14.29%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 2.5028 - accuracy: 0.2857\n",
            "Validation score CBOW 50D: 28.57%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 6.9835 - accuracy: 0.0000e+00\n",
            "Test score CBOW 50D: 0.00%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 2.2893 - accuracy: 0.2857\n",
            "Validation score CBOW 50D: 28.57%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 5.7673 - accuracy: 0.1429\n",
            "Test score CBOW 50D: 14.29%\n",
            "1/1 [==============================] - 0s 942us/step - loss: 2.6939 - accuracy: 0.1429\n",
            "Validation score CBOW 50D: 14.29%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 5.7428 - accuracy: 0.1429\n",
            "Test score CBOW 50D: 14.29%\n",
            "1/1 [==============================] - 0s 961us/step - loss: 2.4026 - accuracy: 0.2857\n",
            "Validation score CBOW 50D: 28.57%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 6.0620 - accuracy: 0.0000e+00\n",
            "Test score CBOW 50D: 0.00%\n",
            "1/1 [==============================] - 0s 1000us/step - loss: 2.8216 - accuracy: 0.1429\n",
            "Validation score CBOW 50D: 14.29%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 6.9297 - accuracy: 0.0000e+00\n",
            "Test score CBOW 50D: 0.00%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 2.3155 - accuracy: 0.2857\n",
            "Validation score CBOW 50D: 28.57%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 6.0325 - accuracy: 0.1429\n",
            "Test score CBOW 50D: 14.29%\n",
            "1/1 [==============================] - 0s 995us/step - loss: 2.4516 - accuracy: 0.2857\n",
            "Validation score CBOW 50D: 28.57%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 7.1121 - accuracy: 0.0000e+00\n",
            "Test score CBOW 50D: 0.00%\n",
            "1/1 [==============================] - 0s 952us/step - loss: 2.3683 - accuracy: 0.2857\n",
            "Validation score CBOW 50D: 28.57%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 6.7365 - accuracy: 0.0000e+00\n",
            "Test score CBOW 50D: 0.00%\n",
            "1/1 [==============================] - 0s 954us/step - loss: 2.7488 - accuracy: 0.2857\n",
            "Validation score CBOW 50D: 28.57%\n",
            "Mean test scores:: 5.71%\n",
            "Mean val scores:: 24.29%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 7.1122 - accuracy: 0.0000e+00\n",
            "Test score SkipGram 50D: 0.00%\n",
            "1/1 [==============================] - 0s 959us/step - loss: 2.3311 - accuracy: 0.1429\n",
            "Validation score SkipGram 50D: 14.29%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 6.3884 - accuracy: 0.0000e+00\n",
            "Test score SkipGram 50D: 0.00%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 2.5705 - accuracy: 0.1429\n",
            "Validation score SkipGram 50D: 14.29%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 5.6183 - accuracy: 0.0000e+00\n",
            "Test score SkipGram 50D: 0.00%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 2.6775 - accuracy: 0.1429\n",
            "Validation score SkipGram 50D: 14.29%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 6.7372 - accuracy: 0.0000e+00\n",
            "Test score SkipGram 50D: 0.00%\n",
            "1/1 [==============================] - 0s 952us/step - loss: 2.7506 - accuracy: 0.1429\n",
            "Validation score SkipGram 50D: 14.29%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 7.0895 - accuracy: 0.0000e+00\n",
            "Test score SkipGram 50D: 0.00%\n",
            "1/1 [==============================] - 0s 972us/step - loss: 2.5244 - accuracy: 0.1429\n",
            "Validation score SkipGram 50D: 14.29%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 5.6499 - accuracy: 0.1429\n",
            "Test score SkipGram 50D: 14.29%\n",
            "1/1 [==============================] - 0s 934us/step - loss: 2.1916 - accuracy: 0.4286\n",
            "Validation score SkipGram 50D: 42.86%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 6.5952 - accuracy: 0.0000e+00\n",
            "Test score SkipGram 50D: 0.00%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 2.4960 - accuracy: 0.1429\n",
            "Validation score SkipGram 50D: 14.29%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 5.5344 - accuracy: 0.1429\n",
            "Test score SkipGram 50D: 14.29%\n",
            "1/1 [==============================] - 0s 966us/step - loss: 2.4800 - accuracy: 0.1429\n",
            "Validation score SkipGram 50D: 14.29%\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 7.1039 - accuracy: 0.0000e+00\n",
            "Test score SkipGram 50D: 0.00%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 2.6466 - accuracy: 0.1429\n",
            "Validation score SkipGram 50D: 14.29%\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 5.8226 - accuracy: 0.1429\n",
            "Test score SkipGram 50D: 14.29%\n",
            "1/1 [==============================] - 0s 992us/step - loss: 2.2712 - accuracy: 0.2857\n",
            "Validation score SkipGram 50D: 28.57%\n",
            "Mean test scores:: 4.29%\n",
            "Mean val scores:: 18.57%\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "y-osuIZnooif",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "3ae954c6-84cc-4784-f533-7e5c03e7ee83"
      },
      "source": [
        "# Brief summary of the scores\n",
        "print(\"%.2f%% (+/- %.2f%%)\" % (np.mean(cv_scores), np.std(cv_scores)))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "22.23% (+/- 5.49%)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "D5VQi88Og0Of"
      },
      "source": [
        "# Save and upload model\n",
        "\n",
        "model_json = model.to_json()\n",
        "with open('model-lstm-2.json', 'w') as json_file:\n",
        "  json_file.write(model_json)\n",
        "model.save_weights('model-lstm-2.h5')\n",
        "\n",
        "uploaded = drive.CreateFile({'title': 'model-lstm-2.h5'})\n",
        "uploaded.SetContentFile('model-lstm-2.h5')\n",
        "uploaded.Upload()\n",
        "print('Uploaded file with ID {}'.format(uploaded.get('id')))\n",
        "\n",
        "uploaded = drive.CreateFile({'title': 'model-lstm-2.json'})\n",
        "uploaded.SetContentFile('model-lstm-2.json')\n",
        "uploaded.Upload()\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "o2ShTMqnSciS"
      },
      "source": [
        "!ls -la"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hKOtkssWQr7Y"
      },
      "source": [
        "from google.colab import files\n",
        "files.download('model-lstm.h5')"
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}