Ensemble_util_brad.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Ensembale Mode here\n",
    "Combine all the sub-model with Bagging method"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import scipy\n",
    "import json\n",
    "import seaborn as sns\n",
    "from collections import namedtuple\n",
    "from sklearn.base import TransformerMixin\n",
    "from sklearn import preprocessing\n",
    "from sklearn import metrics\n",
    "from sklearn.feature_extraction.text import CountVectorizer, TfidfVectorizer\n",
    "from sklearn.model_selection import train_test_split, learning_curve, StratifiedKFold\n",
    "from sklearn.metrics import roc_auc_score, roc_curve, classification_report, confusion_matrix, plot_confusion_matrix\n",
    "from sklearn.pipeline import make_pipeline, Pipeline\n",
    "import joblib\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Figure the basic value"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_path = 'Choesn_model/'\n",
    "configure_file_path = 'Choesn_model/Configuration.json'\n",
    "\n",
    "TRAINFILE = 'keyword.csv'\n",
    "TESTFILE = 'key_word_test.csv'\n",
    "boolean = 'True'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Define the MODEL object"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Model(object):\n",
    "\n",
    "    def  __init__(self, model, name, test_set, topic, is_preprocess, level):\n",
    "        self.model = model\n",
    "        self.name = name\n",
    "        self.test_set = test_set\n",
    "        self.topic = topic\n",
    "        self.is_preprocess = is_preprocess\n",
    "        self.level = level\n",
    "        self.encoding = None\n",
    "\n",
    "    # for debug use\n",
    "    def show(self):\n",
    "        print(\n",
    "            self.name,'\\t',\n",
    "            self.test_set,'\\t',\n",
    "            self.topic,'\\t',\n",
    "            self.is_preprocess,'\\t',\n",
    "            self.level\n",
    "        )\n",
    "\n",
    "    def predict(self, x):\n",
    "        if self.level == 1:\n",
    "            pred_y = self.model.predict(x)\n",
    "            for i in range(pred_y.shape[0]):\n",
    "                pred_y[i] = self.convert(str(pred_y[i]))\n",
    "            \n",
    "            return pred_y.astype(int)\n",
    "        else:\n",
    "            return self.model.predict(x)\n",
    "        \n",
    "    def predict_proba(self, x):\n",
    "        return self.model.predict_proba(x)\n",
    "\n",
    "    def set_encoding(self, encoding):\n",
    "        self.encoding = encoding\n",
    "\n",
    "    def convert(self, x):\n",
    "        return int(self.encoding[x])\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load the model detail from json figuration file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_configuration_to_model(file_path):\n",
    "    '''\n",
    "    Load the json file and figure the parameter of each model\n",
    "\n",
    "    Return: (Tuple Object) with two sub-list\n",
    "            sub-list_1: Model for layer one (For revelent and irrevelent)\n",
    "            sub-list_2: Model for layer two (For topic decision)\n",
    "    '''\n",
    "\n",
    "    with open(configure_file_path, 'r') as json_fp:\n",
    "        configuration = json.load(json_fp)\n",
    "\n",
    "    layer_1 = []\n",
    "    layer_2 = []\n",
    "    \n",
    "    for model_figure in configuration:\n",
    "\n",
    "        # read the figure\n",
    "        model_file = joblib.load(model_path + model_figure['model_name'])\n",
    "        name = model_figure['model_name']\n",
    "        test_set = model_figure['test_set']\n",
    "        topic = model_figure['topic']\n",
    "        is_preprocess = boolean == model_figure['preprocess']\n",
    "        level = int(model_figure['level'])\n",
    "\n",
    "        # New model object to save those arguments\n",
    "        model = Model(model_file, name, test_set, topic, is_preprocess, level)\n",
    "\n",
    "        # append to model list for futher processing\n",
    "        if level == 1:\n",
    "            model.set_encoding(model_figure['encoding'])\n",
    "            layer_1.append(model)\n",
    "        elif level == 2:\n",
    "            layer_2.append(model)\n",
    "\n",
    "    return layer_1,layer_2\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Prepare the testing data and preprocess vector"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_vector(column_name, special=False):\n",
    "    '''\n",
    "    df                      str: The train df\n",
    "    fit_column              str: The column for vector to fit\n",
    "\n",
    "    Return: (Vectorizer Object)\n",
    "            Vectorizer of current column\n",
    "    '''\n",
    "    train_df = pd.read_csv(TRAINFILE)\n",
    "    if special is not False:\n",
    "        train_df = train_df.dropna()\n",
    "        train_df[special] = train_df[special].apply(lambda x: x.replace('_', ''))\n",
    "\n",
    "    # prepare the tranform vector\n",
    "    vector = TfidfVectorizer().fit(train_df[column_name])\n",
    "\n",
    "    return vector\n",
    "\n",
    "def preprocess(df, column_name_list):\n",
    "    '''\n",
    "    This function to use to prepare all the data for ensemble system running\n",
    "    including RAW data and Vector-preprocess data\n",
    "\n",
    "    Return: (Dict object)\n",
    "            A after preprocessing data dict, it order by column_name_list\n",
    "        \n",
    "            ext:\n",
    "            Input: column_name_list: ['key_word_100', 'article_words']\n",
    "            Output: test_data_dict: test_data_dict['key_word_100'] --> key_word_100\n",
    "                                    test_data_dict['article_words'] -> article_words\n",
    "                                    test_data_dict['raw']          --> original data\n",
    "    '''\n",
    "\n",
    "    test_data_dict = {}\n",
    "\n",
    "    # first add original data\n",
    "    test_data_dict['raw'] = df\n",
    "\n",
    "    vector = get_vector('article_words', special='article_words')\n",
    "    \n",
    "    for column in column_name_list:\n",
    "        en_data = vector.transform(df[column])\n",
    "        test_data_dict[str(column)] = en_data\n",
    "\n",
    "    # for special data, add it by manul\n",
    "    vector = get_vector('key_word_100')\n",
    "    test_data_dict['key_word_100_1'] = vector.transform(df['key_word_100'])\n",
    "\n",
    "    return test_data_dict\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv(TESTFILE)\n",
    "dict_data = preprocess(df, ['key_word_100', 'article_words'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Brad's work"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "svm_ir_clf.joblib \t key_word_100_1 \t ['IRRELEVANT'] \t True \t 1\n",
      "rel_clf_tfidf_NB.joblib \t article_words \t ['IRRELEVANT'] \t False \t 1\n",
      "best_logi.joblib \t article_words \t ['IRRELEVANT'] \t False \t 1\n",
      "tfidf_logi_pipe.joblib \t key_word_50 \t ['ARTS CULTURE ENTERTAINMENT', 'DOMESTIC MARKETS', 'SPORTS', 'SHARE LISTINGS', 'FOREX MARKETS', 'DEFENCE'] \t False \t 2\n",
      "clf_NB_1.joblib \t key_word_50 \t ['SHARE LISTINGS', 'DOMESTIC MARKETS', 'ARTS CULTURE ENTERTAINMENT', 'SPORTS', 'DEFENCE', 'FOREX MARKETS', 'SCIENCE AND TECHNOLOGY'] \t False \t 2\n",
      "clf_NB_2.joblib \t key_word_100 \t ['SHARE LISTINGS', 'DOMESTIC MARKETS', 'SPORTS', 'HEALTH', 'MONEY MARKETS', 'DEFENCE'] \t False \t 2\n",
      "clf_NB_3.joblib \t key_word_50 \t ['BIOGRAPHIES PERSONALITIES PEOPLE'] \t False \t 2\n",
      "svm_proba_1.joblib \t key_word_100 \t ['DOMESTIC MARKETS', 'SHARE LISTINGS', 'ARTS CULTURE ENTERTAINMENT', 'SPORTS', 'DEFENCE', 'FOREX MARKETS', 'BIOGRAPHIES PERSONALITIES PEOPLE'] \t True \t 2\n",
      "svm_proba_2.joblib \t article_words \t ['DOMESTIC MARKETS', 'SPORTS', 'FOREX MARKETS', 'HEALTH'] \t True \t 2\n",
      "1_article_words.model \t article_words \t ['SPORTS'] \t False \t 2\n",
      "1_key_word_100.model \t key_word_100 \t ['SPORTS'] \t False \t 2\n",
      "2_article_words.model \t article_words \t ['SPORTS'] \t False \t 2\n",
      "[None, None, None] [None, None, None, None, None, None, None, None, None]\n"
     ]
    }
   ],
   "source": [
    "model_list_1, model_list_2 = load_configuration_to_model(configure_file_path)\n",
    "print([i.show() for i in model_list_1],[i.show() for i in model_list_2])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'ARTS CULTURE ENTERTAINMENT': 0, 'BIOGRAPHIES PERSONALITIES PEOPLE': 1, 'DEFENCE': 2, 'DOMESTIC MARKETS': 3, 'FOREX MARKETS': 4, 'HEALTH': 5, 'IRRELEVANT': 6, 'MONEY MARKETS': 7, 'SCIENCE AND TECHNOLOGY': 8, 'SHARE LISTINGS': 9, 'SPORTS': 10}\n",
      "{0: 'ARTS CULTURE ENTERTAINMENT', 1: 'BIOGRAPHIES PERSONALITIES PEOPLE', 2: 'DEFENCE', 3: 'DOMESTIC MARKETS', 4: 'FOREX MARKETS', 5: 'HEALTH', 6: 'IRRELEVANT', 7: 'MONEY MARKETS', 8: 'SCIENCE AND TECHNOLOGY', 9: 'SHARE LISTINGS', 10: 'SPORTS'}\n"
     ]
    }
   ],
   "source": [
    "topic_encode = dict()\n",
    "topic_decode = dict()\n",
    "\n",
    "topic_list = list(set(df['topic']))\n",
    "topic_list.sort()\n",
    "for num, topic in enumerate(topic_list):\n",
    "    topic_decode[num] = topic\n",
    "    topic_encode[topic] = num\n",
    "print(topic_encode)\n",
    "print(topic_decode)\n",
    "\n",
    "SAMPLE_SIZE = len(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "Vote = namedtuple(\"Vote\", \"model cate proba\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "def vote_proba(votes, topic_dict):\n",
    "    cate_count = len(topic_dict)\n",
    "\n",
    "    column_names = ['predict', 'mean_proba', 'models', 'probas']\n",
    "    df = pd.DataFrame(columns = column_names)\n",
    "\n",
    "    for num, sample in enumerate(votes): \n",
    "        # vote_list to count vote, init with 11 zeros\n",
    "        vote_list = [[] for _ in range(cate_count)]\n",
    "        vote_probas = [[] for _ in range(cate_count)]\n",
    "        for vote in sample:\n",
    "            vote_list[vote.cate].append(vote.model)\n",
    "            vote_probas[vote.cate].append(vote.proba)\n",
    "        vote_count = [len(cate_list) for cate_list in vote_list]\n",
    "        vote_proba_mean = [0 for _ in range(cate_count)]\n",
    "        for i in range(cate_count):\n",
    "            if len(vote_probas[i]) > 0:\n",
    "                vote_proba_mean[i] = sum(vote_probas[i])/len(vote_probas[i])\n",
    "            \n",
    "        # get category chosen by vote\n",
    "        max_vote = max(vote_count)\n",
    "        \n",
    "        # no vote\n",
    "        if max_vote == 0:\n",
    "            res = [None, None, None, None]\n",
    "        else:\n",
    "            # check count\n",
    "            # if draw vote, check proba\n",
    "            if vote_list.count(max_vote) > 1:\n",
    "                chosen_cate = vote_proba_mean.index(max(vote_proba_mean))\n",
    "\n",
    "            # if not draw\n",
    "            else:\n",
    "                chosen_cate = vote_count.index(max_vote)\n",
    "\n",
    "            # append to result DataFrame\n",
    "            res = [topic_dict[chosen_cate], vote_proba_mean[chosen_cate], vote_list[chosen_cate], vote_probas[chosen_cate]]\n",
    "        new_row = pd.Series(res, index=column_names)\n",
    "        df = df.append(new_row, ignore_index=True)\n",
    "        \n",
    "    return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "def irre_vote(model_list, sample_size):\n",
    "    all_votes = [[] for _ in range(sample_size)]\n",
    "    for clf in model_list:\n",
    "            print(clf.name)\n",
    "            if clf.is_preprocess is True:\n",
    "                test_x = dict_data[clf.test_set]\n",
    "            else:\n",
    "                test_x = dict_data['raw'][clf.test_set]\n",
    "            y_pred = clf.predict(test_x)\n",
    "            y_pred_proba = clf.predict_proba(test_x)\n",
    "            print(y_pred)\n",
    "            for i in range(sample_size):\n",
    "                cate, proba = y_pred[i], y_pred_proba[i]\n",
    "\n",
    "                clf_name = clf.name.split('.')[0]\n",
    "                # create namedtuple (model_name, category, category_proba)\n",
    "                all_votes[i].append(Vote(clf_name, cate, proba.max()))\n",
    "    topic_dict = {0: 'IRRELEVANT', 1: 'RELEVANT'}\n",
    "    return vote_proba(all_votes, topic_dict)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "svm_ir_clf.joblib\n",
      "[0 0 1 0 0 1 0 1 1 0 0 0 1 1 0 1 0 1 0 1 0 0 0 0 1 1 0 0 1 1 1 0 0 1 0 1 0\n",
      " 0 1 0 1 1 0 0 0 0 0 1 0 1 1 0 0 1 1 0 0 0 1 1 0 0 0 1 1 0 0 1 1 1 1 1 1 1\n",
      " 1 1 1 0 0 1 0 1 1 1 0 1 1 1 0 0 0 0 0 1 0 1 1 0 1 1 1 1 0 1 0 0 0 1 1 1 0\n",
      " 0 0 0 0 1 1 1 0 1 1 0 0 0 1 0 1 0 0 1 0 1 0 1 0 0 0 1 0 1 1 0 0 0 1 0 1 0\n",
      " 0 1 0 0 1 0 0 1 1 0 1 0 1 1 1 0 1 1 1 1 0 1 1 1 0 0 0 1 1 1 0 1 0 1 1 0 1\n",
      " 0 0 0 0 0 1 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 1 1 0 0 1 0 0 0 0 0 0 1 0 1 1 0\n",
      " 1 0 1 0 0 0 1 0 0 0 1 1 0 1 1 1 1 1 0 0 1 0 0 1 0 1 0 0 1 1 1 1 1 0 1 0 0\n",
      " 1 1 0 1 0 1 1 0 0 1 1 0 1 1 1 0 1 0 0 0 0 0 1 0 0 0 1 1 0 1 0 1 0 0 0 0 0\n",
      " 0 1 0 1 1 0 0 0 1 0 1 1 1 1 1 0 1 0 0 1 1 1 0 1 1 1 1 1 0 1 0 1 0 1 0 1 0\n",
      " 1 1 1 1 0 0 1 0 1 0 0 1 0 0 1 1 0 1 1 1 1 1 0 1 1 0 1 0 1 1 0 0 1 0 0 1 0\n",
      " 1 0 1 0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 1 0 0 1 1 1 0 1 0 1 0 0 1 1 0 1 0 0 0\n",
      " 0 0 0 1 0 0 0 0 1 0 1 1 1 0 1 0 1 0 1 0 0 1 0 1 0 1 0 0 0 1 0 1 1 0 1 1 1\n",
      " 0 0 1 0 0 0 0 1 0 0 0 0 0 0 1 0 1 0 1 1 1 0 1 0 1 0 0 1 0 0 1 0 1 0 1 0 1\n",
      " 1 0 1 1 1 1 1 0 1 0 1 0 1 1 0 1 1 0 0]\n",
      "rel_clf_tfidf_NB.joblib\n",
      "[0 0 1 0 0 1 0 1 1 1 0 0 1 1 0 1 0 0 0 1 0 1 0 0 1 1 0 1 1 1 1 1 0 1 0 1 0\n",
      " 0 1 0 1 1 0 0 0 0 1 1 0 1 1 0 0 1 1 0 0 0 0 1 1 0 0 1 1 0 0 1 1 1 0 1 1 1\n",
      " 1 1 1 0 0 1 0 1 1 1 0 1 0 1 1 0 0 0 0 1 0 1 1 0 1 1 0 1 0 1 0 0 1 1 0 1 0\n",
      " 0 0 0 0 1 1 1 0 1 1 0 0 0 1 0 1 0 0 1 0 1 0 1 0 0 0 1 0 0 1 0 0 1 1 0 1 0\n",
      " 0 1 0 0 1 0 1 1 1 0 1 0 1 1 1 0 1 0 0 1 0 0 1 1 0 0 0 1 1 1 0 1 0 1 1 0 0\n",
      " 0 0 0 0 0 1 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 1 1 0 0 1 0 0 0 1 0 0 1 0 1 0 1\n",
      " 1 0 1 0 1 0 1 0 0 0 1 1 0 1 1 1 0 1 0 0 1 0 0 1 0 1 0 0 1 1 1 1 1 0 1 1 0\n",
      " 1 1 0 1 0 1 1 0 0 1 0 0 1 1 1 0 1 0 0 0 0 0 1 1 0 0 1 1 0 1 0 1 0 0 0 0 0\n",
      " 0 1 0 1 1 0 0 0 0 0 1 0 1 0 1 0 1 0 0 1 1 1 0 1 1 1 1 1 0 1 0 1 1 1 0 1 1\n",
      " 0 1 1 1 0 0 1 0 1 0 0 1 0 0 1 1 0 1 1 1 1 1 0 1 1 0 1 0 1 1 0 0 1 0 0 1 0\n",
      " 1 0 1 0 1 0 0 1 0 0 0 0 1 0 1 1 0 0 1 0 0 0 1 1 0 1 0 1 0 1 1 1 0 0 0 0 0\n",
      " 0 0 0 1 0 0 0 0 1 0 0 1 1 0 1 0 1 0 1 0 0 1 1 1 0 1 0 0 0 0 1 1 1 0 1 1 1\n",
      " 0 0 1 0 0 0 0 1 0 0 0 1 0 0 1 0 1 0 1 1 1 0 1 0 0 0 0 0 0 1 1 0 1 1 1 0 1\n",
      " 1 0 1 1 1 0 1 0 1 0 1 0 0 1 0 1 1 1 1]\n",
      "best_logi.joblib\n",
      "[0 0 1 0 0 1 0 1 1 0 0 0 1 1 0 1 0 1 0 1 0 0 0 0 1 1 0 0 1 1 1 0 1 1 0 1 0\n",
      " 0 1 0 1 1 0 0 0 0 1 1 0 1 1 0 1 1 1 0 0 0 1 1 0 0 0 0 1 0 0 1 1 1 1 1 1 1\n",
      " 0 1 1 0 0 1 0 1 1 1 0 1 1 1 0 0 0 0 0 1 0 1 1 0 1 1 1 1 0 1 0 0 1 1 1 1 0\n",
      " 0 0 0 0 1 1 1 0 1 1 0 0 0 1 0 1 0 0 1 0 1 0 1 0 0 0 1 0 0 1 0 0 0 1 0 1 0\n",
      " 0 1 0 0 1 0 1 1 1 0 1 0 1 1 1 0 1 1 1 1 0 1 1 1 0 0 0 1 1 1 0 1 0 1 1 0 0\n",
      " 0 0 0 0 0 1 0 1 1 1 1 0 0 0 0 1 1 1 1 0 1 1 1 0 0 1 0 0 0 1 0 0 1 0 1 1 0\n",
      " 1 0 1 0 0 0 1 0 0 0 1 1 1 1 1 1 0 1 0 0 1 0 0 0 0 1 1 0 1 1 1 1 1 0 1 1 1\n",
      " 1 1 0 1 0 1 1 0 0 1 1 0 1 1 1 0 1 0 0 0 0 0 1 0 0 0 1 1 0 1 0 1 0 0 0 0 1\n",
      " 1 1 0 1 1 0 0 0 0 1 1 1 1 0 1 0 1 0 0 1 1 1 0 1 1 1 1 1 0 1 0 1 0 1 0 1 1\n",
      " 1 1 1 1 0 0 1 0 0 0 0 1 0 0 1 1 0 1 1 1 1 1 0 1 1 0 1 0 1 1 0 0 1 0 0 1 0\n",
      " 1 0 1 0 1 0 1 1 0 0 0 0 1 0 0 1 0 0 1 0 0 1 1 1 0 1 0 1 0 0 1 1 0 0 0 0 0\n",
      " 0 0 1 1 0 0 0 0 1 0 0 1 1 0 1 0 1 0 1 0 0 1 1 1 0 1 0 0 0 1 1 1 1 0 1 0 1\n",
      " 0 0 1 0 0 0 0 1 0 0 0 0 0 0 1 0 1 0 1 1 1 0 1 0 0 0 0 1 0 0 1 0 1 1 1 0 1\n",
      " 1 0 1 1 1 1 1 1 1 0 1 0 1 1 0 1 1 1 0]\n"
     ]
    }
   ],
   "source": [
    "irre_df = irre_vote(model_list_1, SAMPLE_SIZE)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "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>predict</th>\n",
       "      <th>mean_proba</th>\n",
       "      <th>models</th>\n",
       "      <th>probas</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.855460</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9415586552948044, 1.0, 0.6248227956728621]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.798582</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9294322124011538, 1.0, 0.46631374746266774]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>RELEVANT</td>\n",
       "      <td>0.809855</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.949771764091339, 1.0, 0.47979271676817054]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.783383</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.8360490328877616, 1.0, 0.5140987736785823]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.984085</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9852943113674895, 1.0, 0.9669618390243637]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.834089</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9740126783506491, 1.0, 0.5282558161717601]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>496</th>\n",
       "      <td>RELEVANT</td>\n",
       "      <td>0.985749</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9639243912490126, 1.0, 0.9933233114016177]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497</th>\n",
       "      <td>RELEVANT</td>\n",
       "      <td>0.893329</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9415672706585443, 1.0, 0.7384200805379728]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>498</th>\n",
       "      <td>RELEVANT</td>\n",
       "      <td>0.677644</td>\n",
       "      <td>[rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[1.0, 0.355288766615221]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>499</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.837794</td>\n",
       "      <td>[svm_ir_clf, best_logi]</td>\n",
       "      <td>[0.8109765254551804, 0.8646112118577693]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>500 rows × 4 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "        predict  mean_proba                                     models  \\\n",
       "0    IRRELEVANT    0.855460  [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "1    IRRELEVANT    0.798582  [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "2      RELEVANT    0.809855  [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "3    IRRELEVANT    0.783383  [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "4    IRRELEVANT    0.984085  [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "..          ...         ...                                        ...   \n",
       "495  IRRELEVANT    0.834089  [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "496    RELEVANT    0.985749  [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "497    RELEVANT    0.893329  [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "498    RELEVANT    0.677644              [rel_clf_tfidf_NB, best_logi]   \n",
       "499  IRRELEVANT    0.837794                    [svm_ir_clf, best_logi]   \n",
       "\n",
       "                                             probas  \n",
       "0     [0.9415586552948044, 1.0, 0.6248227956728621]  \n",
       "1    [0.9294322124011538, 1.0, 0.46631374746266774]  \n",
       "2     [0.949771764091339, 1.0, 0.47979271676817054]  \n",
       "3     [0.8360490328877616, 1.0, 0.5140987736785823]  \n",
       "4     [0.9852943113674895, 1.0, 0.9669618390243637]  \n",
       "..                                              ...  \n",
       "495   [0.9740126783506491, 1.0, 0.5282558161717601]  \n",
       "496   [0.9639243912490126, 1.0, 0.9933233114016177]  \n",
       "497   [0.9415672706585443, 1.0, 0.7384200805379728]  \n",
       "498                        [1.0, 0.355288766615221]  \n",
       "499        [0.8109765254551804, 0.8646112118577693]  \n",
       "\n",
       "[500 rows x 4 columns]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "irre_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "def cate_vote(model_list, sample_size, irre_res, truth):\n",
    "    \n",
    "    column_names = ['label', 'predict', 'mean_proba', 'models', 'probas']\n",
    "    final_df = pd.DataFrame(columns = column_names)\n",
    "    \n",
    "    all_votes = [[] for _ in range(sample_size)]\n",
    "    for clf in model_list:\n",
    "        print(clf.name)\n",
    "        if clf.is_preprocess is True:\n",
    "            test_x = dict_data[clf.test_set]\n",
    "        else:\n",
    "            test_x = dict_data['raw'][clf.test_set]\n",
    "        y_pred = clf.predict(test_x)\n",
    "        y_pred_proba = clf.predict_proba(test_x)\n",
    "        print(y_pred)\n",
    "        for i in range(sample_size):\n",
    "            cate, proba = y_pred[i], y_pred_proba[i]\n",
    "            if topic_decode[cate] in clf.topic:\n",
    "                            \n",
    "                clf_name = clf.name.split('.')[0]\n",
    "\n",
    "                all_votes[i].append(Vote(clf_name, cate, proba.max()))\n",
    "    rela_res = vote_proba(all_votes, topic_decode)\n",
    "    for i in range(sample_size):\n",
    "        if irre_res['predict'].iloc[i] == 'IRRELEVANT':\n",
    "            final_res = [truth[i]] + irre_res.iloc[i].tolist()\n",
    "        else:\n",
    "            final_res = [truth[i]] + rela_res.iloc[i].tolist()\n",
    "        new_row = pd.Series(final_res, index=column_names)\n",
    "        final_df = final_df.append(new_row, ignore_index=True)\n",
    "\n",
    "    return final_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tfidf_logi_pipe.joblib\n",
      "[ 6  6  7  3  6  4  2 10  7  6  6  6 10  6  3  7  6  9  3 10  4  7  6  6\n",
      "  7  1  6  6  4  4  4  7  7  7  6 10  3  2  4  0 10  7  6  6  3  6  7  4\n",
      "  6  4  4  3  7  4  7  0  6  6  2  4  7  9  6  7  4  6  4 10 10  4  7  4\n",
      " 10 10  1  2  4  6  2 10  9  6  7  4  6  7  7  4  7  6  6  6  4  7  6 10\n",
      " 10  6  7  7  9  7  6  0  9  7  2 10  5 10  9  8  4  9  6  2  8  7  6 10\n",
      "  8  6  7  6  4  7  7  2  5 10  3  4  6  4  6  6  6  7  6  3  7  6  3  6\n",
      "  3  6  7  6  6  4  6  9 10  6  9 10 10  6  4  1  5 10 10  6  7  9  9  9\n",
      "  6  6  4  4  6  6  6  7  7  4  6  7  3  4  7  9  7  6  6  9  6  6  7  6\n",
      "  4 10 10  4  3  6  6  6 10  4  0  4  9  2  7  7  7  7  4  9  2  7  9  6\n",
      "  6  4  6  0  2  6  7  6  4  6  9  6  7  6  2  6  4  4  5 10  7 10  2  4\n",
      "  4  6  4  6  6  7  9  4  7  3  7 10  7 10  7  6  0  5  6 10  7  3 10  1\n",
      "  7  7  3  1  7  2  6  4  2 10  6  4  6  6  7  3  2 10  1  7  6  4 10  8\n",
      "  0  1 10  7  6  6  6  3  1  4  6 10 10  6  7  6  7  7  5  6 10  5 10  6\n",
      " 10  6  1  7 10 10  6  7 10  7  4  7  6  4  6  7  6  0  4 10  3  0  7 10\n",
      "  4  6  9  7  6  3  6  9  4  6  6 10 10  6  4  4  7  5  4  7 10 10  6  7\n",
      "  6 10  7  9  6  4  9  7  4  6  7  6  5  6  4  9  7  1  9  6  4  2  4  6\n",
      "  7 10  5  6  4  6  6  4  4  4  2  1  7  7  9  0  7  4  6  2  6  6  3  6\n",
      "  7  3  5  6  6  6  7  7  6  6 10 10  6 10  3  4  9  5  0  6  8  7 10  6\n",
      "  0  4  6  7  5  7  7  1  6 10  4  7  6  9  5  1  9  6  9  0  9  3  2  6\n",
      "  6  6  7  6  4  6  4 10  4  6  7  7  7  6  7  9  6  1  4  6  4  6 10  6\n",
      " 10  8  1  4  7  4  2  1  3  7  9 10  6  3  4  8 10  7  9  6]\n",
      "clf_NB_1.joblib\n",
      "[ 3  2  4  3  4  4  2 10  7  4  4  3 10 10  3  7  2  9  3 10  4  4  3  9\n",
      "  7  1  9  4  4  4  4  4  7  7  7 10  7  3  4  9 10  4  9  7  2  2  7  4\n",
      "  3  7  4  3  7  4  4  9  9  9  2  4  4  9  9  4  4  3  4 10 10  4  7  4\n",
      " 10 10  1  2  4  3  2 10  9  1  7  4  7  7  4  4  4  3  3  4  3  7  6 10\n",
      " 10  9  4  7  9  7  5  0  9  7  2 10  5 10  4  5  4  4  4  2  8  7  2 10\n",
      "  5  9  7  9  4  7  7  8  9 10  3  4  9  4  6  3  9  7  9  4  7  4  3  5\n",
      "  2  6  4  2  4  4  3  9 10  9  9 10 10  4  4  2  5 10 10  9  4  9  9  9\n",
      "  4  2  4  4  2  9  7  7  4  4  4  7  3  4  7  9  4  3  4  9  3  4  7  3\n",
      "  4 10 10  4  3  2  3  7 10  4  0  4  9  2  7  7  4  4  4  9  2  2  9  6\n",
      "  4  4  7  7  2  8  7  1  4  9  4  3  4  9  2  9  4  4  5 10  7 10  2  4\n",
      "  2  9  4  7  3  4  9  4  7  3  4 10  4 10  7  2  4  1  2 10  7  3 10  2\n",
      "  7  7  3  2  7  2  2  4  2 10  2  7  2  7  4  3  9 10  2  4  9  4 10  3\n",
      "  0  9 10  7  7  6  3  3  2  4  2 10 10  3  4  9  7  4  5  7 10  5  1  9\n",
      " 10  7  2  7 10 10  2  7 10  7  4  7  4  4  3  7  7  0  2 10  3  0  7 10\n",
      "  4  9  4  7  9  2  9  4  4  9  5 10 10  4  4  4  7  1  4  7 10 10  4  7\n",
      "  4 10  7  9  7  4  9  4  4  4  7  7  5  4  4  9  7  1  4  7  4  2  4  9\n",
      "  4 10  5  4  4  2  6  4  4  4  3  1  7  4  2  9  7  4  3  2  2  7  3  3\n",
      "  3  3  5  9  4  3  4  4  9  9 10 10  9 10  3  7  2  5  2  4  5  7 10  9\n",
      "  0  4  2  7  5  4  7  1  7 10  4  4  9  9  5  4  9  9  9  0  7  3  2  1\n",
      "  4  3  4  3  4  2  4 10  4  9  7  7  7  3  7  9  9  1  4  9  4  5 10  3\n",
      " 10  8  1  4  7  4  2  1  3  7  9 10  9  3  7  8 10  7  9  4]\n",
      "clf_NB_2.joblib\n",
      "[ 9  6  7  6  6  7  6 10  7  7  9  6 10 10  7  7  6  9  3 10  7  7  7  6\n",
      "  7  1  9  7  7  7  7  4  7  7  7 10  7  6  7  9 10  4  6  7  9  8  7  7\n",
      "  3  7  7  3  7  7  7  9  9  7  2  7  4  9  9  4  7  3  7 10 10  7  7  4\n",
      " 10 10  5  2  4  3  2 10  9  1  7  4  7  7  7  7  7  3  3  6  3  7  6 10\n",
      " 10  7  7  7  9  7  9  0  9  7  2 10  5 10  9  5  7  9  6  2  5  7  6 10\n",
      "  5  3  9  9  7  7  7  6  9 10  3  4  3  7  6  9  9  7  9  3  7  4  3  9\n",
      "  3  6  4  2  9  7  3  6 10  9  9 10 10  6  7  2  5 10 10  9  7  9  9  9\n",
      "  7  2  7  7  2  9  7  7  7  4  4  7  3  7  7  9  7  9  6  9  4  6  7  6\n",
      "  4 10 10  4  7  6  3  7 10  7  1  4  9  2  7  7  7  7  4  9  2  6  9  7\n",
      "  7  7  7  7  4  8  7  2  7  6  7  3  7  9  2  9  7  4  5 10  7 10  2  7\n",
      "  2  9  4  7  7  7  6  4  7  3  4 10  7 10  7  3 10  6  2 10  7  3 10  2\n",
      "  7  7  3  2  7  2  9  7  2 10  5  7  2  7  7  3  9 10  2  7  9  7 10  3\n",
      "  0  3 10  6  6  6  3  3  2  7  2 10 10  3  7  9  7  4  5  7 10  5  8  9\n",
      " 10  9  2  7 10 10  5  7 10  7  4  7  4  7  9  7  7  0  6 10  3  9  7 10\n",
      "  4  9  9  7  9  2  9  9  4  9  9 10 10  7  7  7  7  5  7  7 10 10  9  7\n",
      "  4  6  7  9  7  7  9  4  4  7  7  9  5  7  4  9  7  1  9  6  9  2  7  9\n",
      "  7 10  5  4  4  6  3  6  4  7  6  1  7  7  2  6  7  4  3  2  2  7  3  9\n",
      "  3  3  5  9  7  6  7  7  9  6 10 10  9 10  3  7  9  5  1  9  5  7 10  9\n",
      "  1  7  2  9  5  7  7  1  7 10  4  7  7  9  5  6  9  9  9  0  7  3  2  6\n",
      "  7  6  7  6  7  2  4 10  4  9  7  7  7  3  7  9  9  5  7  9  7  5 10  3\n",
      " 10  5  1  7  7  7  2  5  3  7  9 10  7  3  7  9 10  7  9  7]\n",
      "clf_NB_3.joblib\n",
      "[ 3  0  7  6  6  4  6 10  7  4  4  6 10 10  3  7  6  9  3 10  6  4  7  6\n",
      "  7  1  6  4  7  4  4  4  1  7  7 10  7  6  4  6 10  4  6  7  2  2  7  4\n",
      "  3  7  4  3  7  4  7  6  9  9  2  7  4  9  6  4  4  3  4 10 10  4  7  4\n",
      " 10 10  1  2  4  3  2 10  9  1  7  4  7  7  7  4  4  3  3  6  3  7  6 10\n",
      " 10  7  4  7  9  7  0  0  9  7  2 10  5 10  4  5  4  4  6  2  8  7  6 10\n",
      "  5  9  6  9  4  7  7  2  6 10  3  4  6  4  6  6  9  7  6  3  7  4  3  5\n",
      "  1  6  4  2  4  4  3  6 10  9  9 10 10  6  4  2  5 10 10  6  4  9  9  9\n",
      "  4  6  4  4  2  9  7  7  4  4  2  7  3  4  7  9  4  6  4  9  3  6  7  3\n",
      "  4 10 10  4  6  1  6  9 10  4  0  4  6  2  7  7  6  4  4  9  2  2  9  6\n",
      "  4  4  6  7  2  8  7  1  4  9  4  6  4  6  2  9  4  4  5 10  7 10  2  4\n",
      "  2  9  4  7  6  7  6  4  7  3  4 10  4 10  7  6 10  1  2 10  7  3 10  5\n",
      "  7  7  3  1  7  2  2  4  2 10  6  7  2  7  6  3  9 10  1  7  6  4 10  6\n",
      "  0  6 10  7  6  6  3  3  1  4  1 10 10  3  6  9  7  6  5  7 10  5  1  9\n",
      " 10  6  2  7 10 10  2  7 10  7  4  7  4  4  6  7  7  0  6 10  3  0  7 10\n",
      "  4  6  4  7  6  2  6  6  4  9  6 10 10  7  4  4  7  1  4  7 10 10  6  7\n",
      "  4 10  7  9  6  4  9  6  4  7  7  6  5  4  4  9  7  1  4  6  4  2  4  6\n",
      "  7 10  5  6  4  6  6  6  4  4  6  1  7  7  2  6  7  4  6  2  2  7  3  6\n",
      "  7  3  5  9  4  6  4  4  6  6 10 10  9 10  3  7  6  5  0  6  5  7 10  9\n",
      "  0  4  2  6  5  4  7  1  9 10  4  7  6  9  5  6  9  9  9  0  7  3  2  1\n",
      "  4  6  4  6  4  2  4 10  4  9  7  7  7  3  7  9  9  1  4  6  4  5 10  3\n",
      " 10  5  1  4  7  4  2  1  3  7  6 10  6  3  7  6 10  7  9  4]\n",
      "svm_proba_1.joblib\n",
      "[ 3  6  4  6  6  4  2 10  7  9  9  6 10 10  6  7  6  9  3 10  6  4  7  6\n",
      "  7  1  6  7  4  4  4  4  1  7  6 10  3  6  4  6 10  4  6  8  2  1  7  4\n",
      "  3  4  4  3  7  4  4  6  9  6  2  7  4  9  9  4  4  6  4 10 10  4  7  4\n",
      " 10 10  1  2  4  6  2 10  9  1  7  4  6  7  7  4  4  6  3  6  6  4  6 10\n",
      " 10  6  4  7  9  7  6  1  9  7  2 10  5 10  4  6  4  9  6  2  8  7  2 10\n",
      "  5  9  7  6  4  4  4  6  6 10  6  4  6  4  6  6  6  7  6  3  7  6  3  6\n",
      "  1  6  4  2  4  4  3  9 10  9  9 10 10  9  4  6  5 10 10  6  4  9  9  6\n",
      "  1  2  4  7  2  6  7  4  4  4  2  7  6  4  7  9  4  6  6  9  3  6  7  3\n",
      "  4 10 10  4  3  1  6  9 10  4  0  4  6  2  7  4  4  4  4  9  2  2  9  6\n",
      "  4  4  6  7  2  8  7  1  4  6  9  6  4  9  2  9  4  4  5 10  7 10  2  4\n",
      "  2  9  4  7  6  7  9  4  7  3  4 10  3 10  7  6 10  1  2 10  7  6 10  1\n",
      "  7  7  3  1  7  2  2  4  2 10  5  4  2  7  4  3  9 10  2  4  6  4 10  6\n",
      "  0  6 10  7  6  6  3  3  1  4  1 10 10  3  7  9  7  4  5  6 10  5 10  9\n",
      " 10  6  2  7 10 10  6  7 10  7  4  7  4  4  6  7  7  0  6 10  3  0  7 10\n",
      "  4  6  9  7  6  2  6  9  4  6  6 10 10  9  4  4  7  1  4  7 10 10  6  7\n",
      "  4 10  7  9  6  4  9  6  4  6  7  6  5  4  4  9  7  1  9  6  4  2  4  9\n",
      "  4 10  5  9  4  2  6  2  4  4  9  1  7  7  2  6  7  4  3  2  2  8  3  6\n",
      "  6  3  5  9  4  6  7  7  6  6 10 10  9 10  3  4  6  5  2  6  5  7 10  6\n",
      "  0  4  2  7  5  7  7  1  9 10  4  4  9  9  5  1  9  9  9  0  9  3  2  1\n",
      "  6  3  4  6  4  6  4 10  4  9  7  7  4  6  7  9  6  6  4  6  4  5 10  6\n",
      " 10  5  1  4  7  4  2  1  3  7  6 10  6  3  7  6 10  7  9  4]\n",
      "svm_proba_2.joblib\n",
      "[ 6  6  4  6  6  4  6 10  7  6  9  6 10 10  6  7  6  9  3 10  6  4  6  6\n",
      "  7  0  6  6  4  4  4  4  6  7  6 10  6  6  4  6 10  6  6  6  6  6  7  4\n",
      "  6  7  4  6  7  4  4  6  6  6  2  7  4  9  6  4  4  6  6 10 10  4  6  4\n",
      " 10 10  1  2  4  6  6 10  6  1  7  4  6  7  7  4  6  6  6  6  6  7  6 10\n",
      " 10  6  4  7  9  7  6  6  6  7  2 10  5 10  6  6  6  6  6  2  8  7  6 10\n",
      "  5  6  6  6  4  6  4  6  6 10  6  4  6  4  6  6  6  7  6  3  4  6  6  6\n",
      "  1  6  4  6  6  4  6  6 10  6  6 10 10  6  4  6  5 10 10  6  4  9  6  6\n",
      "  1  2  4  7  6  6  6  4  4  4  6  7  6  4  4  6  4  6  6  6  6  6  7  3\n",
      "  4 10 10  4  6  6  6  6 10  4  0  4  6  2  7  4  6  7  4  9  2  6  9  6\n",
      "  6  4  6  7  6  8  7  1  4  6  6  6  4  6  2  6  4  4  5 10  7 10  2  4\n",
      "  2  6  4  6  6  7  6  4  6  3  4 10  4 10  7  6 10  1  2 10  7  6 10  6\n",
      "  7  7  3  1  7  2  6  4  2 10  6  4  6  6  6  3  6 10  6  6  6  4 10  6\n",
      "  0  6 10  6  6  6  6  3  1  4  6 10 10  6  6  6  7  6  5  6 10  5 10  6\n",
      " 10  6  2  7 10 10  6  7 10  7  4  7  6  4  6  7  7  0  6 10  5  6  7 10\n",
      "  4  6  6  7  6  2  6  6  4  6  6 10 10  6  4  4  7  1  4  6 10 10  6  7\n",
      "  6 10  7  9  6  4  9  6  4  6  7  6  5  6  4  6  7  1  6  6  4  2  6  6\n",
      "  4 10  5  6  4  2  6  2  6  4  6  1  6  7  6  6  4  4  6  6  6  6  3  6\n",
      "  6  3  5  6  6  6  6  7  6  6 10 10  6 10  3  4  6  5  6  6  5  7 10  6\n",
      "  0  6  6  6  5  7  7  1  6 10  4  4  6  9  5  6  6  6  9  0  6  3  2  6\n",
      "  6  3  4  6  4  6  4 10  4  6  7  7  6  6  6  9  6  6  4  6  4  5 10  6\n",
      " 10  5  1  4  7  4  2  1  3  7  6 10  6  3  7  6 10  7  6  6]\n",
      "1_article_words.model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 6  6  7  6  6  4  7 10  7  6  6  6 10 10  3  6  6  9  3 10  4  6  6  6\n",
      "  4  6  6  9  6  7  7  6  6  7  6 10  3  6  7  9  6  4  6  6  6  6  7  4\n",
      "  6  4  7  6  6  7  4  6  6  6  0  7  6  6  6  9  7  6  6 10 10  7  7  6\n",
      " 10 10  6  2  4  6  6 10  9 10  7  7  6  7  7  4  4  6  6  0  6  7  6 10\n",
      " 10  6  7  7  9  7  6  6  6  6 10 10  6 10  4  6  6  9  6  6  6  7  4 10\n",
      "  5  6  1  6  4  6  4  6  6 10  6  7  7  7  6  6  6  7  6  6  4  6  6  6\n",
      "  6  6  7  6  9  4  6  6 10  6  9 10 10  9  4  6  5 10 10  6  4  9  9  4\n",
      "  6  2  7  7  6  6  6  7  4  4  2  7  3  4  4  6  6  6  6  6  6  6  7  6\n",
      "  4 10 10  9  6  8  6  6 10  7  0  4  6  2  7  4  6  6  4  6  2  1  9  6\n",
      "  6  7  6  6  6  6  6  6  7  6  6  6  4  6  2  6  7  7  5 10  7 10  2  4\n",
      "  6  6  4  7  6  8  9  7  6  6  7  6  4 10 10  6  6  0  6 10  7  6 10  6\n",
      "  6  7  6  6  7  5  6  4  6 10  6  7  6  6  6  6  6 10  1 10  6  4 10  6\n",
      "  6  6 10  6  6  6  6  6 10  4  3 10 10  6  7  6  6  4  3  6 10  1 10  7\n",
      " 10  6  6  7 10  6  0  7 10  7  6  4  6  4  6  7  4  0  6 10  6  6  7 10\n",
      "  4  6  6  7  6  2  6  6  7  6  6 10 10  6  7  7  7  5  7  6 10 10  6  7\n",
      "  6 10  7  6  6  4  6  6  6  6  7  6  5  6  4  6  7  6  7  6  4  2  6  6\n",
      "  6 10  5  6  4  2  6  6  7  4  6 10  7  7  6  6  6  4  6  1  0  6  6  6\n",
      "  6  2  5  6  6  6  6  7  6  6 10 10  6 10  6  4  6  1  6  6  0  6 10  6\n",
      "  0  6  6  6  5  6  7  6  6 10  7  4  6  6  6  6  4  6  6  6  6  6  6  6\n",
      "  6  6  7  6  7  6  7 10  7  6  7  6  4  6  1  9  6  6  4  6  4  6 10  6\n",
      " 10  8 10  7  6  4  2  1  6  7  6 10  6  6  7  2 10 10  6  6]\n",
      "1_key_word_100.model\n",
      "[ 6  6  7  6  6  4  6 10  7  6  6  6 10 10  6  7 10  6  6 10  7  9  6  6\n",
      "  7  9  7  6  7  4  7  6  6  7  6 10  3  6  7  6 10  4  6  6  6  1  7  4\n",
      "  6  7  4  6  6  7  4  6  6  6  2  4  6  6  6  9  7  6  7 10 10  7  7  6\n",
      " 10 10  1  2  7  6  6 10  9 10  7  4  6  7  7  4  7  6  6  6  6  6  6 10\n",
      " 10  6  4  6  9  7  6  6  6  7  6 10  6 10  6  5  6  6  6  2  6  7  6 10\n",
      "  5  6  7  6  4  6  4  6  6 10  6  7  6  7  6  6  6  7  6  6  4  6  6  6\n",
      "  2  1  4  6  6  4  6  6 10  6  6 10 10  6  7  6  6 10 10  6  6  9  9  4\n",
      "  6  2  7  7  2  6  6  7  7  4  6  7  6  4  7  6  7  6  6  6  6  6  7  3\n",
      "  4 10 10  6  6  6  6  7 10  7  0  4  4  6  7  4  6  6  4  6  6  6  9  6\n",
      "  6  4  6  6  6  6  7  6  7  6  6  6  7  6  2  6  7  7  6 10  7 10  6  6\n",
      "  6  6  7  7  6  6  7  4  5  6  7 10  7 10 10  6  6  0  6 10  7  6 10  6\n",
      "  7  7  6  6  7  2  6  7  6 10  6  7  6  6  6  6  6 10  1 10  6  6 10  6\n",
      "  1  6 10  6  6  6  6  3 10  4  2 10 10  6  6  6  7  7  6  7 10  5  6  6\n",
      " 10  6  6  7 10  6  6  7  6  4  7  7  6  7  6  7  4  0  6 10  3  7  7 10\n",
      "  4  6  6  7  6  6  6  6  4  6  4 10  6  6  6  4  7  5  7  6 10 10  6  7\n",
      "  6 10  7  6  6  7  6  6  7  6  7  6  5  6  4  6  6  0  6  6  7  2  6  6\n",
      "  6 10  6  6  4  2  6  6  6  4  6  4  6  7  3  4  7  4  2  5  5  6  6  6\n",
      "  6  6 10  6  6  6  6  7  6  6 10  6  6  6  6  4  6  6  6  6  6  4 10  6\n",
      "  1  6  6  6  5  6  7  5  6 10  4  7  6  6  6  6  6  6  6  1  6  6  6  6\n",
      "  6  6  6 10  4  6  4 10  7  6  6  6  4  6  6  9  6  6  4  6  7 10 10  6\n",
      " 10  8  1  7  7  4  2  1  3  7  6 10  6  6  7  6 10 10  6  6]\n",
      "2_article_words.model\n",
      "[ 6  6  7  6  6  7  7 10  7  6  6  6 10 10  6  7  6  9  7 10  6  6  6  6\n",
      "  4  6  6  6  4  4  7  6  6  7  6 10  6  7  7  6 10  4  6  6  6  5  6  4\n",
      "  6  4  4  6  6  4  4  6  6  6  2  7  7  9  6  9  4  6  6 10  2  7  6  4\n",
      " 10 10  1  8  6  6  6 10  9 10  7  7 10  7  7  4  6  8  6  4  7  7  6 10\n",
      " 10  6  7  7  6  7  1  6  6  4  2 10  5 10 10  8  4  9  8  6  3  7  7 10\n",
      "  5  6  6  6  4  6  7  6  6 10 10  7  6  7  6  6  6  7  6  6  4  6  7  6\n",
      "  1  6  7  0  6  7  6  9  1  6  6 10 10  6  7  6  8 10 10  6  4  6  9  6\n",
      "  7  4  6  6  7  6  6  7  4  4  4  4  6  4  4  6  7  6  6  9  6  6  7  6\n",
      "  7 10 10  7  7  5  7  6 10  7  0  7  6  6  7  4  6  6  7  6  2  6  6  6\n",
      "  4  4  6  6  2  6  7  6  7  1  3  6  4  9  6  6  7  6  6  6  6 10 10  4\n",
      "  1  9  7  6  6  7  6  7  6  6  7 10  6 10  7  6  9  1  6 10  7  6 10  6\n",
      "  7  7  6  5  7  2  6  7  2 10  6  4  0  1  6  6  6 10  6  6  6  4 10  6\n",
      "  0  6 10  6  6  6  6  0 10  4  6 10  6  6  6  6  7  6  6  6 10  6  1  6\n",
      " 10  6  6  7 10  6  1  7 10  7  4  4  4  7  6  7  4  0  4 10  3  5  7  6\n",
      "  4  6  6  7  6  1  6  6  4  6  6 10 10  6  7  7  7  5  7  6 10 10  6  7\n",
      "  6  6  7  9  6  4  6  6  7  6  7  5  6  6  7  9  6  6  6  6  6  1  6  6\n",
      "  6 10  6  6  4  6  6  4  7  4  6  6  7  4 10  6  4  7  6  2  1  6  6  6\n",
      "  6  6  6  6  6  6  6  7  6  6 10 10  6 10  6  4  6  1  1 10  8  6 10  6\n",
      " 10  6  9  6  5  6  7  1  6 10  4  7  6  9  5  6  6  6  9  0  6  6  6  1\n",
      "  6  6  4 10  7  1  7 10  7  6  7  6  4  6  7  9  6  6  7  9  4 10 10  6\n",
      " 10  5  6  4  6  4  2 10  3  7  1 10  6  6  7  6 10  7  9  6]\n"
     ]
    }
   ],
   "source": [
    "res_df = cate_vote(model_list_2, SAMPLE_SIZE, irre_df, df['topic'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "# if there is None, uncomment this\n",
    "# res_df.loc[res_df.predict.isnull()] = 'IRRELEVANT'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classification Report:\n",
      "                                  precision    recall  f1-score   support\n",
      "\n",
      "      ARTS CULTURE ENTERTAINMENT       0.38      1.00      0.55         3\n",
      "BIOGRAPHIES PERSONALITIES PEOPLE       0.78      0.47      0.58        15\n",
      "                         DEFENCE       0.90      0.69      0.78        13\n",
      "                DOMESTIC MARKETS       0.50      0.50      0.50         2\n",
      "                   FOREX MARKETS       0.47      0.90      0.61        48\n",
      "                          HEALTH       0.69      0.64      0.67        14\n",
      "                      IRRELEVANT       0.91      0.87      0.89       266\n",
      "                   MONEY MARKETS       0.63      0.35      0.45        69\n",
      "          SCIENCE AND TECHNOLOGY       0.00      0.00      0.00         3\n",
      "                  SHARE LISTINGS       0.56      0.71      0.63         7\n",
      "                          SPORTS       0.92      1.00      0.96        60\n",
      "\n",
      "                        accuracy                           0.79       500\n",
      "                       macro avg       0.61      0.65      0.60       500\n",
      "                    weighted avg       0.81      0.79      0.78       500\n",
      "\n",
      "[[  3   0   0   0   0   0   0   0   0   0   0]\n",
      " [  3   7   0   0   0   0   5   0   0   0   0]\n",
      " [  0   0   9   0   0   0   4   0   0   0   0]\n",
      " [  0   0   0   1   0   0   1   0   0   0   0]\n",
      " [  0   0   0   0  43   0   2   2   0   1   0]\n",
      " [  0   0   0   1   0   9   4   0   0   0   0]\n",
      " [  0   2   1   0  10   3 232  12   0   3   3]\n",
      " [  1   0   0   0  39   0   3  24   0   0   2]\n",
      " [  1   0   0   0   0   1   1   0   0   0   0]\n",
      " [  0   0   0   0   0   0   2   0   0   5   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0  60]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 936x504 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def report(test_y, y_pred):\n",
    "    global topic_list\n",
    "\n",
    "    print(f\"Classification Report:\\n{classification_report(test_y, y_pred)}\")\n",
    "#     print(\"The scores for each topic is:\")\n",
    "#     scores_for_topics(df_test,topics,model,le)\n",
    "    conf_matrix = confusion_matrix(test_y, y_pred)\n",
    "    print(conf_matrix)\n",
    "    fig1 = plt.figure(figsize=(13,7))\n",
    "    sns.heatmap(conf_matrix,\n",
    "    #             square=True,\n",
    "                annot=True, # show numbers in each cell\n",
    "                fmt='d', # set number format to integer in each cell\n",
    "                yticklabels=topic_list,\n",
    "#                 xticklabels=topic_list,\n",
    "                cmap=\"Blues\",\n",
    "    #             linecolor=\"k\",\n",
    "                linewidths=.1,\n",
    "               )\n",
    "    plt.title(\n",
    "              f\"Confusion Matrix on Test Set | \" \n",
    "              f\"Classifier: Ensemble Model\", \n",
    "              fontsize=14)\n",
    "    plt.xlabel(\"Actual: False positives for y != x\", fontsize=12)\n",
    "    plt.ylabel(\"Prediction: False negatives for x != y\", fontsize=12)\n",
    "    plt.show()\n",
    "\n",
    "report(res_df['label'], res_df['predict'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "DEFENCE accuarcy is:\t\t\t\t 0.6923076923076923\n",
      "DOMESTIC MARKETS accuarcy is:\t\t\t\t 0.5\n",
      "MONEY MARKETS accuarcy is:\t\t\t\t 0.34782608695652173\n",
      "FOREX MARKETS accuarcy is:\t\t\t\t 0.8958333333333334\n",
      "SPORTS accuarcy is:\t\t\t\t 1.0\n",
      "SHARE LISTINGS accuarcy is:\t\t\t\t 0.7142857142857143\n",
      "BIOGRAPHIES PERSONALITIES PEOPLE accuarcy is:\t\t\t\t 0.4666666666666667\n",
      "HEALTH accuarcy is:\t\t\t\t 0.6428571428571429\n",
      "ARTS CULTURE ENTERTAINMENT accuarcy is:\t\t\t\t 1.0\n",
      "SCIENCE AND TECHNOLOGY accuarcy is:\t\t\t\t 0.0\n",
      "IRRELEVANT accuarcy is:\t\t\t\t 0.8721804511278195\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def evaluate(res_df, y_true_label, y_pred_label):\n",
    "    '''\n",
    "    Here is for the evaluate the ensamble model\n",
    "\n",
    "    Input: (DataFrame Object) Should be result of the prediction\n",
    "    Output: \n",
    "    '''\n",
    "\n",
    "    report = []\n",
    "    topic_list = list(set(res_df[y_true_label]))\n",
    "    for topic in topic_list:\n",
    "\n",
    "        # prepare the record\n",
    "        topic_report = {}\n",
    "        topic_report['name'] = str(topic)\n",
    "\n",
    "        # prepare the evaluate data\n",
    "        test_df = res_df[res_df[y_true_label] == topic]\n",
    "\n",
    "        #evaluate each part\n",
    "        topic_report['f1_score'] = metrics.f1_score(test_df[y_true_label], test_df[y_pred_label], average='macro')\n",
    "        topic_report['accuarcy'] = metrics.accuracy_score(test_df[y_true_label], test_df[y_pred_label])\n",
    "        topic_report['recall_score'] = metrics.recall_score(test_df[y_true_label], test_df[y_pred_label], average='macro')\n",
    "        print(topic,'accuarcy is:\\t\\t\\t\\t', topic_report['accuarcy'])\n",
    "\n",
    "        # append t total report for further\n",
    "        report.append(topic_report)\n",
    "        \n",
    "\n",
    "    # sort the report for plt\n",
    "    report.sort(reverse=True, key=lambda x: x['accuarcy'])\n",
    "\n",
    "    #plt.style.use('ggplot')\n",
    "    figure = plt.figure(figsize=(12,6))\n",
    "    #plt.xticks(rotation=90)\n",
    "    plt.title('Accuarcy in each topic')\n",
    "    plt.barh([i['name'] for i in report], [j['accuarcy'] for j in report])\n",
    "\n",
    "    plt.show()\n",
    "\n",
    "evaluate(res_df, 'label', 'predict')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "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>label</th>\n",
       "      <th>predict</th>\n",
       "      <th>mean_proba</th>\n",
       "      <th>models</th>\n",
       "      <th>probas</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.85546</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9415586552948044, 1.0, 0.6248227956728621]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.798582</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9294322124011538, 1.0, 0.46631374746266774]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.783383</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.8360490328877616, 1.0, 0.5140987736785823]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.984085</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9852943113674895, 1.0, 0.9669618390243637]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.857133</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.8658562226628614, 1.0, 0.705542322929795]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>488</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.858489</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB]</td>\n",
       "      <td>[0.7169770939662491, 1.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>490</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.821487</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9277460669417049, 1.0, 0.536715232310648]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>492</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.838036</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9402443410480955, 1.0, 0.5738634061656198]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.834089</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9740126783506491, 1.0, 0.5282558161717601]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>499</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.837794</td>\n",
       "      <td>[svm_ir_clf, best_logi]</td>\n",
       "      <td>[0.8109765254551804, 0.8646112118577693]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>266 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "          label     predict mean_proba  \\\n",
       "0    IRRELEVANT  IRRELEVANT    0.85546   \n",
       "1    IRRELEVANT  IRRELEVANT   0.798582   \n",
       "3    IRRELEVANT  IRRELEVANT   0.783383   \n",
       "4    IRRELEVANT  IRRELEVANT   0.984085   \n",
       "6    IRRELEVANT  IRRELEVANT   0.857133   \n",
       "..          ...         ...        ...   \n",
       "488  IRRELEVANT  IRRELEVANT   0.858489   \n",
       "490  IRRELEVANT  IRRELEVANT   0.821487   \n",
       "492  IRRELEVANT  IRRELEVANT   0.838036   \n",
       "495  IRRELEVANT  IRRELEVANT   0.834089   \n",
       "499  IRRELEVANT  IRRELEVANT   0.837794   \n",
       "\n",
       "                                        models  \\\n",
       "0    [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "1    [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "3    [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "4    [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "6    [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "..                                         ...   \n",
       "488             [svm_ir_clf, rel_clf_tfidf_NB]   \n",
       "490  [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "492  [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "495  [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "499                    [svm_ir_clf, best_logi]   \n",
       "\n",
       "                                             probas  \n",
       "0     [0.9415586552948044, 1.0, 0.6248227956728621]  \n",
       "1    [0.9294322124011538, 1.0, 0.46631374746266774]  \n",
       "3     [0.8360490328877616, 1.0, 0.5140987736785823]  \n",
       "4     [0.9852943113674895, 1.0, 0.9669618390243637]  \n",
       "6      [0.8658562226628614, 1.0, 0.705542322929795]  \n",
       "..                                              ...  \n",
       "488                       [0.7169770939662491, 1.0]  \n",
       "490    [0.9277460669417049, 1.0, 0.536715232310648]  \n",
       "492   [0.9402443410480955, 1.0, 0.5738634061656198]  \n",
       "495   [0.9740126783506491, 1.0, 0.5282558161717601]  \n",
       "499        [0.8109765254551804, 0.8646112118577693]  \n",
       "\n",
       "[266 rows x 5 columns]"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "temp_rs_df = res_df[res_df['label'] == 'IRRELEVANT']\n",
    "temp_rs_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "For not corrcet prediction\n"
     ]
    },
    {
     "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>label</th>\n",
       "      <th>predict</th>\n",
       "      <th>mean_proba</th>\n",
       "      <th>models</th>\n",
       "      <th>probas</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>80</th>\n",
       "      <td>BIOGRAPHIES PERSONALITIES PEOPLE</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.812736</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.8987773645522399, 1.0, 0.539429660328727]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>267</th>\n",
       "      <td>BIOGRAPHIES PERSONALITIES PEOPLE</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.908493</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9261901825678165, 1.0, 0.7992895558077636]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>282</th>\n",
       "      <td>BIOGRAPHIES PERSONALITIES PEOPLE</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.731552</td>\n",
       "      <td>[svm_ir_clf, best_logi]</td>\n",
       "      <td>[0.8199555314762236, 0.6431475892186023]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>288</th>\n",
       "      <td>BIOGRAPHIES PERSONALITIES PEOPLE</td>\n",
       "      <td>ARTS CULTURE ENTERTAINMENT</td>\n",
       "      <td>0.994053</td>\n",
       "      <td>[tfidf_logi_pipe, clf_NB_1, svm_proba_1]</td>\n",
       "      <td>[0.999995684598029, 1.0, 0.982163650771372]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>329</th>\n",
       "      <td>BIOGRAPHIES PERSONALITIES PEOPLE</td>\n",
       "      <td>ARTS CULTURE ENTERTAINMENT</td>\n",
       "      <td>0.999147</td>\n",
       "      <td>[tfidf_logi_pipe, clf_NB_1, svm_proba_1]</td>\n",
       "      <td>[0.9976895900514323, 1.0, 0.9997517201501492]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>432</th>\n",
       "      <td>BIOGRAPHIES PERSONALITIES PEOPLE</td>\n",
       "      <td>ARTS CULTURE ENTERTAINMENT</td>\n",
       "      <td>0.994979</td>\n",
       "      <td>[tfidf_logi_pipe, clf_NB_1, svm_proba_1]</td>\n",
       "      <td>[0.9999999996591098, 1.0, 0.9849355928393099]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>455</th>\n",
       "      <td>BIOGRAPHIES PERSONALITIES PEOPLE</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.579564</td>\n",
       "      <td>[svm_ir_clf, best_logi]</td>\n",
       "      <td>[0.6243131545534658, 0.5348144300028544]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>482</th>\n",
       "      <td>BIOGRAPHIES PERSONALITIES PEOPLE</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.72819</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.5213612083655645, 1.0, 0.6632099125938061]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                label                     predict mean_proba  \\\n",
       "80   BIOGRAPHIES PERSONALITIES PEOPLE                  IRRELEVANT   0.812736   \n",
       "267  BIOGRAPHIES PERSONALITIES PEOPLE                  IRRELEVANT   0.908493   \n",
       "282  BIOGRAPHIES PERSONALITIES PEOPLE                  IRRELEVANT   0.731552   \n",
       "288  BIOGRAPHIES PERSONALITIES PEOPLE  ARTS CULTURE ENTERTAINMENT   0.994053   \n",
       "329  BIOGRAPHIES PERSONALITIES PEOPLE  ARTS CULTURE ENTERTAINMENT   0.999147   \n",
       "432  BIOGRAPHIES PERSONALITIES PEOPLE  ARTS CULTURE ENTERTAINMENT   0.994979   \n",
       "455  BIOGRAPHIES PERSONALITIES PEOPLE                  IRRELEVANT   0.579564   \n",
       "482  BIOGRAPHIES PERSONALITIES PEOPLE                  IRRELEVANT    0.72819   \n",
       "\n",
       "                                        models  \\\n",
       "80   [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "267  [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "282                    [svm_ir_clf, best_logi]   \n",
       "288   [tfidf_logi_pipe, clf_NB_1, svm_proba_1]   \n",
       "329   [tfidf_logi_pipe, clf_NB_1, svm_proba_1]   \n",
       "432   [tfidf_logi_pipe, clf_NB_1, svm_proba_1]   \n",
       "455                    [svm_ir_clf, best_logi]   \n",
       "482  [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "\n",
       "                                            probas  \n",
       "80    [0.8987773645522399, 1.0, 0.539429660328727]  \n",
       "267  [0.9261901825678165, 1.0, 0.7992895558077636]  \n",
       "282       [0.8199555314762236, 0.6431475892186023]  \n",
       "288    [0.999995684598029, 1.0, 0.982163650771372]  \n",
       "329  [0.9976895900514323, 1.0, 0.9997517201501492]  \n",
       "432  [0.9999999996591098, 1.0, 0.9849355928393099]  \n",
       "455       [0.6243131545534658, 0.5348144300028544]  \n",
       "482  [0.5213612083655645, 1.0, 0.6632099125938061]  "
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "temp_rs_df = res_df[res_df['label'] == 'BIOGRAPHIES PERSONALITIES PEOPLE']\n",
    "temp_rs_df\n",
    "print('For not corrcet prediction')\n",
    "no_correct = temp_rs_df[temp_rs_df['predict'] != 'BIOGRAPHIES PERSONALITIES PEOPLE']\n",
    "no_correct\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "svm_ir_clf was vote for 5 times\n",
      "rel_clf_tfidf_NB was vote for 3 times\n",
      "best_logi was vote for 5 times\n",
      "tfidf_logi_pipe was vote for 3 times\n",
      "clf_NB_1 was vote for 3 times\n",
      "svm_proba_1 was vote for 3 times\n"
     ]
    }
   ],
   "source": [
    "# calculate the incorrect model vote time:\n",
    "vote_time = {}\n",
    "for index, row in no_correct.iterrows():\n",
    "    for model in row['models']:\n",
    "        if model not in vote_time:\n",
    "            vote_time[model] = 1\n",
    "        else:\n",
    "            vote_time[model] += 1\n",
    "\n",
    "for model in vote_time:\n",
    "    print(model, 'was vote for', vote_time[model], 'times')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "# check json topics\n",
    "for model in model_list_1 + model_list_2:\n",
    "    for topic in model.topic:\n",
    "        if topic not in topic_list:\n",
    "            print(topic)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "svm_ir_clf.joblib\n",
      "[0 0 1 0 0 1 0 1 1 0 0 0 1 1 0 1 0 1 0 1 0 0 0 0 1 1 0 0 1 1 1 0 0 1 0 1 0\n",
      " 0 1 0 1 1 0 0 0 0 0 1 0 1 1 0 0 1 1 0 0 0 1 1 0 0 0 1 1 0 0 1 1 1 1 1 1 1\n",
      " 1 1 1 0 0 1 0 1 1 1 0 1 1 1 0 0 0 0 0 1 0 1 1 0 1 1 1 1 0 1 0 0 0 1 1 1 0\n",
      " 0 0 0 0 1 1 1 0 1 1 0 0 0 1 0 1 0 0 1 0 1 0 1 0 0 0 1 0 1 1 0 0 0 1 0 1 0\n",
      " 0 1 0 0 1 0 0 1 1 0 1 0 1 1 1 0 1 1 1 1 0 1 1 1 0 0 0 1 1 1 0 1 0 1 1 0 1\n",
      " 0 0 0 0 0 1 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 1 1 0 0 1 0 0 0 0 0 0 1 0 1 1 0\n",
      " 1 0 1 0 0 0 1 0 0 0 1 1 0 1 1 1 1 1 0 0 1 0 0 1 0 1 0 0 1 1 1 1 1 0 1 0 0\n",
      " 1 1 0 1 0 1 1 0 0 1 1 0 1 1 1 0 1 0 0 0 0 0 1 0 0 0 1 1 0 1 0 1 0 0 0 0 0\n",
      " 0 1 0 1 1 0 0 0 1 0 1 1 1 1 1 0 1 0 0 1 1 1 0 1 1 1 1 1 0 1 0 1 0 1 0 1 0\n",
      " 1 1 1 1 0 0 1 0 1 0 0 1 0 0 1 1 0 1 1 1 1 1 0 1 1 0 1 0 1 1 0 0 1 0 0 1 0\n",
      " 1 0 1 0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 1 0 0 1 1 1 0 1 0 1 0 0 1 1 0 1 0 0 0\n",
      " 0 0 0 1 0 0 0 0 1 0 1 1 1 0 1 0 1 0 1 0 0 1 0 1 0 1 0 0 0 1 0 1 1 0 1 1 1\n",
      " 0 0 1 0 0 0 0 1 0 0 0 0 0 0 1 0 1 0 1 1 1 0 1 0 1 0 0 1 0 0 1 0 1 0 1 0 1\n",
      " 1 0 1 1 1 1 1 0 1 0 1 0 1 1 0 1 1 0 0]\n",
      "rel_clf_tfidf_NB.joblib\n",
      "[0 0 1 0 0 1 0 1 1 1 0 0 1 1 0 1 0 0 0 1 0 1 0 0 1 1 0 1 1 1 1 1 0 1 0 1 0\n",
      " 0 1 0 1 1 0 0 0 0 1 1 0 1 1 0 0 1 1 0 0 0 0 1 1 0 0 1 1 0 0 1 1 1 0 1 1 1\n",
      " 1 1 1 0 0 1 0 1 1 1 0 1 0 1 1 0 0 0 0 1 0 1 1 0 1 1 0 1 0 1 0 0 1 1 0 1 0\n",
      " 0 0 0 0 1 1 1 0 1 1 0 0 0 1 0 1 0 0 1 0 1 0 1 0 0 0 1 0 0 1 0 0 1 1 0 1 0\n",
      " 0 1 0 0 1 0 1 1 1 0 1 0 1 1 1 0 1 0 0 1 0 0 1 1 0 0 0 1 1 1 0 1 0 1 1 0 0\n",
      " 0 0 0 0 0 1 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 1 1 0 0 1 0 0 0 1 0 0 1 0 1 0 1\n",
      " 1 0 1 0 1 0 1 0 0 0 1 1 0 1 1 1 0 1 0 0 1 0 0 1 0 1 0 0 1 1 1 1 1 0 1 1 0\n",
      " 1 1 0 1 0 1 1 0 0 1 0 0 1 1 1 0 1 0 0 0 0 0 1 1 0 0 1 1 0 1 0 1 0 0 0 0 0\n",
      " 0 1 0 1 1 0 0 0 0 0 1 0 1 0 1 0 1 0 0 1 1 1 0 1 1 1 1 1 0 1 0 1 1 1 0 1 1\n",
      " 0 1 1 1 0 0 1 0 1 0 0 1 0 0 1 1 0 1 1 1 1 1 0 1 1 0 1 0 1 1 0 0 1 0 0 1 0\n",
      " 1 0 1 0 1 0 0 1 0 0 0 0 1 0 1 1 0 0 1 0 0 0 1 1 0 1 0 1 0 1 1 1 0 0 0 0 0\n",
      " 0 0 0 1 0 0 0 0 1 0 0 1 1 0 1 0 1 0 1 0 0 1 1 1 0 1 0 0 0 0 1 1 1 0 1 1 1\n",
      " 0 0 1 0 0 0 0 1 0 0 0 1 0 0 1 0 1 0 1 1 1 0 1 0 0 0 0 0 0 1 1 0 1 1 1 0 1\n",
      " 1 0 1 1 1 0 1 0 1 0 1 0 0 1 0 1 1 1 1]\n",
      "best_logi.joblib\n",
      "[0 0 1 0 0 1 0 1 1 0 0 0 1 1 0 1 0 1 0 1 0 0 0 0 1 1 0 0 1 1 1 0 1 1 0 1 0\n",
      " 0 1 0 1 1 0 0 0 0 1 1 0 1 1 0 1 1 1 0 0 0 1 1 0 0 0 0 1 0 0 1 1 1 1 1 1 1\n",
      " 0 1 1 0 0 1 0 1 1 1 0 1 1 1 0 0 0 0 0 1 0 1 1 0 1 1 1 1 0 1 0 0 1 1 1 1 0\n",
      " 0 0 0 0 1 1 1 0 1 1 0 0 0 1 0 1 0 0 1 0 1 0 1 0 0 0 1 0 0 1 0 0 0 1 0 1 0\n",
      " 0 1 0 0 1 0 1 1 1 0 1 0 1 1 1 0 1 1 1 1 0 1 1 1 0 0 0 1 1 1 0 1 0 1 1 0 0\n",
      " 0 0 0 0 0 1 0 1 1 1 1 0 0 0 0 1 1 1 1 0 1 1 1 0 0 1 0 0 0 1 0 0 1 0 1 1 0\n",
      " 1 0 1 0 0 0 1 0 0 0 1 1 1 1 1 1 0 1 0 0 1 0 0 0 0 1 1 0 1 1 1 1 1 0 1 1 1\n",
      " 1 1 0 1 0 1 1 0 0 1 1 0 1 1 1 0 1 0 0 0 0 0 1 0 0 0 1 1 0 1 0 1 0 0 0 0 1\n",
      " 1 1 0 1 1 0 0 0 0 1 1 1 1 0 1 0 1 0 0 1 1 1 0 1 1 1 1 1 0 1 0 1 0 1 0 1 1\n",
      " 1 1 1 1 0 0 1 0 0 0 0 1 0 0 1 1 0 1 1 1 1 1 0 1 1 0 1 0 1 1 0 0 1 0 0 1 0\n",
      " 1 0 1 0 1 0 1 1 0 0 0 0 1 0 0 1 0 0 1 0 0 1 1 1 0 1 0 1 0 0 1 1 0 0 0 0 0\n",
      " 0 0 1 1 0 0 0 0 1 0 0 1 1 0 1 0 1 0 1 0 0 1 1 1 0 1 0 0 0 1 1 1 1 0 1 0 1\n",
      " 0 0 1 0 0 0 0 1 0 0 0 0 0 0 1 0 1 0 1 1 1 0 1 0 0 0 0 1 0 0 1 0 1 1 1 0 1\n",
      " 1 0 1 1 1 1 1 1 1 0 1 0 1 1 0 1 1 1 0]\n",
      "tfidf_logi_pipe.joblib\n",
      "[ 6  6  7  3  6  4  2 10  7  6  6  6 10  6  3  7  6  9  3 10  4  7  6  6\n",
      "  7  1  6  6  4  4  4  7  7  7  6 10  3  2  4  0 10  7  6  6  3  6  7  4\n",
      "  6  4  4  3  7  4  7  0  6  6  2  4  7  9  6  7  4  6  4 10 10  4  7  4\n",
      " 10 10  1  2  4  6  2 10  9  6  7  4  6  7  7  4  7  6  6  6  4  7  6 10\n",
      " 10  6  7  7  9  7  6  0  9  7  2 10  5 10  9  8  4  9  6  2  8  7  6 10\n",
      "  8  6  7  6  4  7  7  2  5 10  3  4  6  4  6  6  6  7  6  3  7  6  3  6\n",
      "  3  6  7  6  6  4  6  9 10  6  9 10 10  6  4  1  5 10 10  6  7  9  9  9\n",
      "  6  6  4  4  6  6  6  7  7  4  6  7  3  4  7  9  7  6  6  9  6  6  7  6\n",
      "  4 10 10  4  3  6  6  6 10  4  0  4  9  2  7  7  7  7  4  9  2  7  9  6\n",
      "  6  4  6  0  2  6  7  6  4  6  9  6  7  6  2  6  4  4  5 10  7 10  2  4\n",
      "  4  6  4  6  6  7  9  4  7  3  7 10  7 10  7  6  0  5  6 10  7  3 10  1\n",
      "  7  7  3  1  7  2  6  4  2 10  6  4  6  6  7  3  2 10  1  7  6  4 10  8\n",
      "  0  1 10  7  6  6  6  3  1  4  6 10 10  6  7  6  7  7  5  6 10  5 10  6\n",
      " 10  6  1  7 10 10  6  7 10  7  4  7  6  4  6  7  6  0  4 10  3  0  7 10\n",
      "  4  6  9  7  6  3  6  9  4  6  6 10 10  6  4  4  7  5  4  7 10 10  6  7\n",
      "  6 10  7  9  6  4  9  7  4  6  7  6  5  6  4  9  7  1  9  6  4  2  4  6\n",
      "  7 10  5  6  4  6  6  4  4  4  2  1  7  7  9  0  7  4  6  2  6  6  3  6\n",
      "  7  3  5  6  6  6  7  7  6  6 10 10  6 10  3  4  9  5  0  6  8  7 10  6\n",
      "  0  4  6  7  5  7  7  1  6 10  4  7  6  9  5  1  9  6  9  0  9  3  2  6\n",
      "  6  6  7  6  4  6  4 10  4  6  7  7  7  6  7  9  6  1  4  6  4  6 10  6\n",
      " 10  8  1  4  7  4  2  1  3  7  9 10  6  3  4  8 10  7  9  6]\n",
      "clf_NB_1.joblib\n",
      "[ 3  2  4  3  4  4  2 10  7  4  4  3 10 10  3  7  2  9  3 10  4  4  3  9\n",
      "  7  1  9  4  4  4  4  4  7  7  7 10  7  3  4  9 10  4  9  7  2  2  7  4\n",
      "  3  7  4  3  7  4  4  9  9  9  2  4  4  9  9  4  4  3  4 10 10  4  7  4\n",
      " 10 10  1  2  4  3  2 10  9  1  7  4  7  7  4  4  4  3  3  4  3  7  6 10\n",
      " 10  9  4  7  9  7  5  0  9  7  2 10  5 10  4  5  4  4  4  2  8  7  2 10\n",
      "  5  9  7  9  4  7  7  8  9 10  3  4  9  4  6  3  9  7  9  4  7  4  3  5\n",
      "  2  6  4  2  4  4  3  9 10  9  9 10 10  4  4  2  5 10 10  9  4  9  9  9\n",
      "  4  2  4  4  2  9  7  7  4  4  4  7  3  4  7  9  4  3  4  9  3  4  7  3\n",
      "  4 10 10  4  3  2  3  7 10  4  0  4  9  2  7  7  4  4  4  9  2  2  9  6\n",
      "  4  4  7  7  2  8  7  1  4  9  4  3  4  9  2  9  4  4  5 10  7 10  2  4\n",
      "  2  9  4  7  3  4  9  4  7  3  4 10  4 10  7  2  4  1  2 10  7  3 10  2\n",
      "  7  7  3  2  7  2  2  4  2 10  2  7  2  7  4  3  9 10  2  4  9  4 10  3\n",
      "  0  9 10  7  7  6  3  3  2  4  2 10 10  3  4  9  7  4  5  7 10  5  1  9\n",
      " 10  7  2  7 10 10  2  7 10  7  4  7  4  4  3  7  7  0  2 10  3  0  7 10\n",
      "  4  9  4  7  9  2  9  4  4  9  5 10 10  4  4  4  7  1  4  7 10 10  4  7\n",
      "  4 10  7  9  7  4  9  4  4  4  7  7  5  4  4  9  7  1  4  7  4  2  4  9\n",
      "  4 10  5  4  4  2  6  4  4  4  3  1  7  4  2  9  7  4  3  2  2  7  3  3\n",
      "  3  3  5  9  4  3  4  4  9  9 10 10  9 10  3  7  2  5  2  4  5  7 10  9\n",
      "  0  4  2  7  5  4  7  1  7 10  4  4  9  9  5  4  9  9  9  0  7  3  2  1\n",
      "  4  3  4  3  4  2  4 10  4  9  7  7  7  3  7  9  9  1  4  9  4  5 10  3\n",
      " 10  8  1  4  7  4  2  1  3  7  9 10  9  3  7  8 10  7  9  4]\n",
      "clf_NB_2.joblib\n",
      "[ 9  6  7  6  6  7  6 10  7  7  9  6 10 10  7  7  6  9  3 10  7  7  7  6\n",
      "  7  1  9  7  7  7  7  4  7  7  7 10  7  6  7  9 10  4  6  7  9  8  7  7\n",
      "  3  7  7  3  7  7  7  9  9  7  2  7  4  9  9  4  7  3  7 10 10  7  7  4\n",
      " 10 10  5  2  4  3  2 10  9  1  7  4  7  7  7  7  7  3  3  6  3  7  6 10\n",
      " 10  7  7  7  9  7  9  0  9  7  2 10  5 10  9  5  7  9  6  2  5  7  6 10\n",
      "  5  3  9  9  7  7  7  6  9 10  3  4  3  7  6  9  9  7  9  3  7  4  3  9\n",
      "  3  6  4  2  9  7  3  6 10  9  9 10 10  6  7  2  5 10 10  9  7  9  9  9\n",
      "  7  2  7  7  2  9  7  7  7  4  4  7  3  7  7  9  7  9  6  9  4  6  7  6\n",
      "  4 10 10  4  7  6  3  7 10  7  1  4  9  2  7  7  7  7  4  9  2  6  9  7\n",
      "  7  7  7  7  4  8  7  2  7  6  7  3  7  9  2  9  7  4  5 10  7 10  2  7\n",
      "  2  9  4  7  7  7  6  4  7  3  4 10  7 10  7  3 10  6  2 10  7  3 10  2\n",
      "  7  7  3  2  7  2  9  7  2 10  5  7  2  7  7  3  9 10  2  7  9  7 10  3\n",
      "  0  3 10  6  6  6  3  3  2  7  2 10 10  3  7  9  7  4  5  7 10  5  8  9\n",
      " 10  9  2  7 10 10  5  7 10  7  4  7  4  7  9  7  7  0  6 10  3  9  7 10\n",
      "  4  9  9  7  9  2  9  9  4  9  9 10 10  7  7  7  7  5  7  7 10 10  9  7\n",
      "  4  6  7  9  7  7  9  4  4  7  7  9  5  7  4  9  7  1  9  6  9  2  7  9\n",
      "  7 10  5  4  4  6  3  6  4  7  6  1  7  7  2  6  7  4  3  2  2  7  3  9\n",
      "  3  3  5  9  7  6  7  7  9  6 10 10  9 10  3  7  9  5  1  9  5  7 10  9\n",
      "  1  7  2  9  5  7  7  1  7 10  4  7  7  9  5  6  9  9  9  0  7  3  2  6\n",
      "  7  6  7  6  7  2  4 10  4  9  7  7  7  3  7  9  9  5  7  9  7  5 10  3\n",
      " 10  5  1  7  7  7  2  5  3  7  9 10  7  3  7  9 10  7  9  7]\n",
      "clf_NB_3.joblib\n",
      "[ 3  0  7  6  6  4  6 10  7  4  4  6 10 10  3  7  6  9  3 10  6  4  7  6\n",
      "  7  1  6  4  7  4  4  4  1  7  7 10  7  6  4  6 10  4  6  7  2  2  7  4\n",
      "  3  7  4  3  7  4  7  6  9  9  2  7  4  9  6  4  4  3  4 10 10  4  7  4\n",
      " 10 10  1  2  4  3  2 10  9  1  7  4  7  7  7  4  4  3  3  6  3  7  6 10\n",
      " 10  7  4  7  9  7  0  0  9  7  2 10  5 10  4  5  4  4  6  2  8  7  6 10\n",
      "  5  9  6  9  4  7  7  2  6 10  3  4  6  4  6  6  9  7  6  3  7  4  3  5\n",
      "  1  6  4  2  4  4  3  6 10  9  9 10 10  6  4  2  5 10 10  6  4  9  9  9\n",
      "  4  6  4  4  2  9  7  7  4  4  2  7  3  4  7  9  4  6  4  9  3  6  7  3\n",
      "  4 10 10  4  6  1  6  9 10  4  0  4  6  2  7  7  6  4  4  9  2  2  9  6\n",
      "  4  4  6  7  2  8  7  1  4  9  4  6  4  6  2  9  4  4  5 10  7 10  2  4\n",
      "  2  9  4  7  6  7  6  4  7  3  4 10  4 10  7  6 10  1  2 10  7  3 10  5\n",
      "  7  7  3  1  7  2  2  4  2 10  6  7  2  7  6  3  9 10  1  7  6  4 10  6\n",
      "  0  6 10  7  6  6  3  3  1  4  1 10 10  3  6  9  7  6  5  7 10  5  1  9\n",
      " 10  6  2  7 10 10  2  7 10  7  4  7  4  4  6  7  7  0  6 10  3  0  7 10\n",
      "  4  6  4  7  6  2  6  6  4  9  6 10 10  7  4  4  7  1  4  7 10 10  6  7\n",
      "  4 10  7  9  6  4  9  6  4  7  7  6  5  4  4  9  7  1  4  6  4  2  4  6\n",
      "  7 10  5  6  4  6  6  6  4  4  6  1  7  7  2  6  7  4  6  2  2  7  3  6\n",
      "  7  3  5  9  4  6  4  4  6  6 10 10  9 10  3  7  6  5  0  6  5  7 10  9\n",
      "  0  4  2  6  5  4  7  1  9 10  4  7  6  9  5  6  9  9  9  0  7  3  2  1\n",
      "  4  6  4  6  4  2  4 10  4  9  7  7  7  3  7  9  9  1  4  6  4  5 10  3\n",
      " 10  5  1  4  7  4  2  1  3  7  6 10  6  3  7  6 10  7  9  4]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "svm_proba_1.joblib\n",
      "[ 3  6  4  6  6  4  2 10  7  9  9  6 10 10  6  7  6  9  3 10  6  4  7  6\n",
      "  7  1  6  7  4  4  4  4  1  7  6 10  3  6  4  6 10  4  6  8  2  1  7  4\n",
      "  3  4  4  3  7  4  4  6  9  6  2  7  4  9  9  4  4  6  4 10 10  4  7  4\n",
      " 10 10  1  2  4  6  2 10  9  1  7  4  6  7  7  4  4  6  3  6  6  4  6 10\n",
      " 10  6  4  7  9  7  6  1  9  7  2 10  5 10  4  6  4  9  6  2  8  7  2 10\n",
      "  5  9  7  6  4  4  4  6  6 10  6  4  6  4  6  6  6  7  6  3  7  6  3  6\n",
      "  1  6  4  2  4  4  3  9 10  9  9 10 10  9  4  6  5 10 10  6  4  9  9  6\n",
      "  1  2  4  7  2  6  7  4  4  4  2  7  6  4  7  9  4  6  6  9  3  6  7  3\n",
      "  4 10 10  4  3  1  6  9 10  4  0  4  6  2  7  4  4  4  4  9  2  2  9  6\n",
      "  4  4  6  7  2  8  7  1  4  6  9  6  4  9  2  9  4  4  5 10  7 10  2  4\n",
      "  2  9  4  7  6  7  9  4  7  3  4 10  3 10  7  6 10  1  2 10  7  6 10  1\n",
      "  7  7  3  1  7  2  2  4  2 10  5  4  2  7  4  3  9 10  2  4  6  4 10  6\n",
      "  0  6 10  7  6  6  3  3  1  4  1 10 10  3  7  9  7  4  5  6 10  5 10  9\n",
      " 10  6  2  7 10 10  6  7 10  7  4  7  4  4  6  7  7  0  6 10  3  0  7 10\n",
      "  4  6  9  7  6  2  6  9  4  6  6 10 10  9  4  4  7  1  4  7 10 10  6  7\n",
      "  4 10  7  9  6  4  9  6  4  6  7  6  5  4  4  9  7  1  9  6  4  2  4  9\n",
      "  4 10  5  9  4  2  6  2  4  4  9  1  7  7  2  6  7  4  3  2  2  8  3  6\n",
      "  6  3  5  9  4  6  7  7  6  6 10 10  9 10  3  4  6  5  2  6  5  7 10  6\n",
      "  0  4  2  7  5  7  7  1  9 10  4  4  9  9  5  1  9  9  9  0  9  3  2  1\n",
      "  6  3  4  6  4  6  4 10  4  9  7  7  4  6  7  9  6  6  4  6  4  5 10  6\n",
      " 10  5  1  4  7  4  2  1  3  7  6 10  6  3  7  6 10  7  9  4]\n",
      "svm_proba_2.joblib\n",
      "[ 6  6  4  6  6  4  6 10  7  6  9  6 10 10  6  7  6  9  3 10  6  4  6  6\n",
      "  7  0  6  6  4  4  4  4  6  7  6 10  6  6  4  6 10  6  6  6  6  6  7  4\n",
      "  6  7  4  6  7  4  4  6  6  6  2  7  4  9  6  4  4  6  6 10 10  4  6  4\n",
      " 10 10  1  2  4  6  6 10  6  1  7  4  6  7  7  4  6  6  6  6  6  7  6 10\n",
      " 10  6  4  7  9  7  6  6  6  7  2 10  5 10  6  6  6  6  6  2  8  7  6 10\n",
      "  5  6  6  6  4  6  4  6  6 10  6  4  6  4  6  6  6  7  6  3  4  6  6  6\n",
      "  1  6  4  6  6  4  6  6 10  6  6 10 10  6  4  6  5 10 10  6  4  9  6  6\n",
      "  1  2  4  7  6  6  6  4  4  4  6  7  6  4  4  6  4  6  6  6  6  6  7  3\n",
      "  4 10 10  4  6  6  6  6 10  4  0  4  6  2  7  4  6  7  4  9  2  6  9  6\n",
      "  6  4  6  7  6  8  7  1  4  6  6  6  4  6  2  6  4  4  5 10  7 10  2  4\n",
      "  2  6  4  6  6  7  6  4  6  3  4 10  4 10  7  6 10  1  2 10  7  6 10  6\n",
      "  7  7  3  1  7  2  6  4  2 10  6  4  6  6  6  3  6 10  6  6  6  4 10  6\n",
      "  0  6 10  6  6  6  6  3  1  4  6 10 10  6  6  6  7  6  5  6 10  5 10  6\n",
      " 10  6  2  7 10 10  6  7 10  7  4  7  6  4  6  7  7  0  6 10  5  6  7 10\n",
      "  4  6  6  7  6  2  6  6  4  6  6 10 10  6  4  4  7  1  4  6 10 10  6  7\n",
      "  6 10  7  9  6  4  9  6  4  6  7  6  5  6  4  6  7  1  6  6  4  2  6  6\n",
      "  4 10  5  6  4  2  6  2  6  4  6  1  6  7  6  6  4  4  6  6  6  6  3  6\n",
      "  6  3  5  6  6  6  6  7  6  6 10 10  6 10  3  4  6  5  6  6  5  7 10  6\n",
      "  0  6  6  6  5  7  7  1  6 10  4  4  6  9  5  6  6  6  9  0  6  3  2  6\n",
      "  6  3  4  6  4  6  4 10  4  6  7  7  6  6  6  9  6  6  4  6  4  5 10  6\n",
      " 10  5  1  4  7  4  2  1  3  7  6 10  6  3  7  6 10  7  6  6]\n",
      "1_article_words.model\n",
      "[ 6  6  7  6  6  4  7 10  7  6  6  6 10 10  3  6  6  9  3 10  4  6  6  6\n",
      "  4  6  6  9  6  7  7  6  6  7  6 10  3  6  7  9  6  4  6  6  6  6  7  4\n",
      "  6  4  7  6  6  7  4  6  6  6  0  7  6  6  6  9  7  6  6 10 10  7  7  6\n",
      " 10 10  6  2  4  6  6 10  9 10  7  7  6  7  7  4  4  6  6  0  6  7  6 10\n",
      " 10  6  7  7  9  7  6  6  6  6 10 10  6 10  4  6  6  9  6  6  6  7  4 10\n",
      "  5  6  1  6  4  6  4  6  6 10  6  7  7  7  6  6  6  7  6  6  4  6  6  6\n",
      "  6  6  7  6  9  4  6  6 10  6  9 10 10  9  4  6  5 10 10  6  4  9  9  4\n",
      "  6  2  7  7  6  6  6  7  4  4  2  7  3  4  4  6  6  6  6  6  6  6  7  6\n",
      "  4 10 10  9  6  8  6  6 10  7  0  4  6  2  7  4  6  6  4  6  2  1  9  6\n",
      "  6  7  6  6  6  6  6  6  7  6  6  6  4  6  2  6  7  7  5 10  7 10  2  4\n",
      "  6  6  4  7  6  8  9  7  6  6  7  6  4 10 10  6  6  0  6 10  7  6 10  6\n",
      "  6  7  6  6  7  5  6  4  6 10  6  7  6  6  6  6  6 10  1 10  6  4 10  6\n",
      "  6  6 10  6  6  6  6  6 10  4  3 10 10  6  7  6  6  4  3  6 10  1 10  7\n",
      " 10  6  6  7 10  6  0  7 10  7  6  4  6  4  6  7  4  0  6 10  6  6  7 10\n",
      "  4  6  6  7  6  2  6  6  7  6  6 10 10  6  7  7  7  5  7  6 10 10  6  7\n",
      "  6 10  7  6  6  4  6  6  6  6  7  6  5  6  4  6  7  6  7  6  4  2  6  6\n",
      "  6 10  5  6  4  2  6  6  7  4  6 10  7  7  6  6  6  4  6  1  0  6  6  6\n",
      "  6  2  5  6  6  6  6  7  6  6 10 10  6 10  6  4  6  1  6  6  0  6 10  6\n",
      "  0  6  6  6  5  6  7  6  6 10  7  4  6  6  6  6  4  6  6  6  6  6  6  6\n",
      "  6  6  7  6  7  6  7 10  7  6  7  6  4  6  1  9  6  6  4  6  4  6 10  6\n",
      " 10  8 10  7  6  4  2  1  6  7  6 10  6  6  7  2 10 10  6  6]\n",
      "1_key_word_100.model\n",
      "[ 6  6  7  6  6  4  6 10  7  6  6  6 10 10  6  7 10  6  6 10  7  9  6  6\n",
      "  7  9  7  6  7  4  7  6  6  7  6 10  3  6  7  6 10  4  6  6  6  1  7  4\n",
      "  6  7  4  6  6  7  4  6  6  6  2  4  6  6  6  9  7  6  7 10 10  7  7  6\n",
      " 10 10  1  2  7  6  6 10  9 10  7  4  6  7  7  4  7  6  6  6  6  6  6 10\n",
      " 10  6  4  6  9  7  6  6  6  7  6 10  6 10  6  5  6  6  6  2  6  7  6 10\n",
      "  5  6  7  6  4  6  4  6  6 10  6  7  6  7  6  6  6  7  6  6  4  6  6  6\n",
      "  2  1  4  6  6  4  6  6 10  6  6 10 10  6  7  6  6 10 10  6  6  9  9  4\n",
      "  6  2  7  7  2  6  6  7  7  4  6  7  6  4  7  6  7  6  6  6  6  6  7  3\n",
      "  4 10 10  6  6  6  6  7 10  7  0  4  4  6  7  4  6  6  4  6  6  6  9  6\n",
      "  6  4  6  6  6  6  7  6  7  6  6  6  7  6  2  6  7  7  6 10  7 10  6  6\n",
      "  6  6  7  7  6  6  7  4  5  6  7 10  7 10 10  6  6  0  6 10  7  6 10  6\n",
      "  7  7  6  6  7  2  6  7  6 10  6  7  6  6  6  6  6 10  1 10  6  6 10  6\n",
      "  1  6 10  6  6  6  6  3 10  4  2 10 10  6  6  6  7  7  6  7 10  5  6  6\n",
      " 10  6  6  7 10  6  6  7  6  4  7  7  6  7  6  7  4  0  6 10  3  7  7 10\n",
      "  4  6  6  7  6  6  6  6  4  6  4 10  6  6  6  4  7  5  7  6 10 10  6  7\n",
      "  6 10  7  6  6  7  6  6  7  6  7  6  5  6  4  6  6  0  6  6  7  2  6  6\n",
      "  6 10  6  6  4  2  6  6  6  4  6  4  6  7  3  4  7  4  2  5  5  6  6  6\n",
      "  6  6 10  6  6  6  6  7  6  6 10  6  6  6  6  4  6  6  6  6  6  4 10  6\n",
      "  1  6  6  6  5  6  7  5  6 10  4  7  6  6  6  6  6  6  6  1  6  6  6  6\n",
      "  6  6  6 10  4  6  4 10  7  6  6  6  4  6  6  9  6  6  4  6  7 10 10  6\n",
      " 10  8  1  7  7  4  2  1  3  7  6 10  6  6  7  6 10 10  6  6]\n",
      "2_article_words.model\n",
      "[ 6  6  7  6  6  7  7 10  7  6  6  6 10 10  6  7  6  9  7 10  6  6  6  6\n",
      "  4  6  6  6  4  4  7  6  6  7  6 10  6  7  7  6 10  4  6  6  6  5  6  4\n",
      "  6  4  4  6  6  4  4  6  6  6  2  7  7  9  6  9  4  6  6 10  2  7  6  4\n",
      " 10 10  1  8  6  6  6 10  9 10  7  7 10  7  7  4  6  8  6  4  7  7  6 10\n",
      " 10  6  7  7  6  7  1  6  6  4  2 10  5 10 10  8  4  9  8  6  3  7  7 10\n",
      "  5  6  6  6  4  6  7  6  6 10 10  7  6  7  6  6  6  7  6  6  4  6  7  6\n",
      "  1  6  7  0  6  7  6  9  1  6  6 10 10  6  7  6  8 10 10  6  4  6  9  6\n",
      "  7  4  6  6  7  6  6  7  4  4  4  4  6  4  4  6  7  6  6  9  6  6  7  6\n",
      "  7 10 10  7  7  5  7  6 10  7  0  7  6  6  7  4  6  6  7  6  2  6  6  6\n",
      "  4  4  6  6  2  6  7  6  7  1  3  6  4  9  6  6  7  6  6  6  6 10 10  4\n",
      "  1  9  7  6  6  7  6  7  6  6  7 10  6 10  7  6  9  1  6 10  7  6 10  6\n",
      "  7  7  6  5  7  2  6  7  2 10  6  4  0  1  6  6  6 10  6  6  6  4 10  6\n",
      "  0  6 10  6  6  6  6  0 10  4  6 10  6  6  6  6  7  6  6  6 10  6  1  6\n",
      " 10  6  6  7 10  6  1  7 10  7  4  4  4  7  6  7  4  0  4 10  3  5  7  6\n",
      "  4  6  6  7  6  1  6  6  4  6  6 10 10  6  7  7  7  5  7  6 10 10  6  7\n",
      "  6  6  7  9  6  4  6  6  7  6  7  5  6  6  7  9  6  6  6  6  6  1  6  6\n",
      "  6 10  6  6  4  6  6  4  7  4  6  6  7  4 10  6  4  7  6  2  1  6  6  6\n",
      "  6  6  6  6  6  6  6  7  6  6 10 10  6 10  6  4  6  1  1 10  8  6 10  6\n",
      " 10  6  9  6  5  6  7  1  6 10  4  7  6  9  5  6  6  6  9  0  6  6  6  1\n",
      "  6  6  4 10  7  1  7 10  7  6  7  6  4  6  7  9  6  6  7  9  4 10 10  6\n",
      " 10  5  6  4  6  4  2 10  3  7  1 10  6  6  7  6 10  7  9  6]\n"
     ]
    }
   ],
   "source": [
    "# ADD_WEIGHT_LIST = [0 for _ in topic_list]\n",
    "# ADD_WEIGHT_LIST[topic_encode['BIOGRAPHIES PERSONALITIES PEOPLE']] = 2\n",
    "# ADD_WEIGHT_LIST[topic_encode['HEALTH']] = 1\n",
    "# ADD_WEIGHT_LIST[topic_encode['MONEY MARKETS']] = 3\n",
    "\n",
    "topic_count = {}\n",
    "for topic in topic_list:\n",
    "    topic_count[topic] = 0\n",
    "    \n",
    "for model in model_list_1 + model_list_2:\n",
    "    for topic in model.topic:\n",
    "        topic_count[topic] += 1\n",
    "ADD_WEIGHT_LIST = [topic_count[max(topic_count)]-i for i in topic_count.values()]\n",
    "ADD_WEIGHT_LIST\n",
    "\n",
    "def vote_proba_2(votes, topic_dict):\n",
    "    cate_count = len(topic_dict)\n",
    "\n",
    "    column_names = ['predict', 'mean_proba', 'models', 'probas']\n",
    "    df = pd.DataFrame(columns = column_names)\n",
    "\n",
    "    for num, sample in enumerate(votes): \n",
    "        # vote_list to count vote, init with 11 zeros\n",
    "        vote_list = [[] for _ in range(cate_count)]\n",
    "        vote_probas = [[] for _ in range(cate_count)]\n",
    "        for vote in sample:\n",
    "            vote_list[vote.cate].append(vote.model)\n",
    "            vote_probas[vote.cate].append(vote.proba)\n",
    "        vote_count = [len(cate_list) for cate_list in vote_list]\n",
    "        vote_proba_mean = [0 for _ in range(cate_count)]\n",
    "        for i in range(cate_count):\n",
    "            if len(vote_probas[i]) > 0:\n",
    "                vote_proba_mean[i] = sum(vote_probas[i])/len(vote_probas[i])\n",
    "                \n",
    "        ###### add weight ######\n",
    "        for i in range(len(vote_count)):\n",
    "            if vote_count[i] > 0:\n",
    "                vote_count[i] += ADD_WEIGHT_LIST[i]\n",
    "        ########################\n",
    "            \n",
    "        # get category chosen by vote\n",
    "        max_vote = max(vote_count)\n",
    "        \n",
    "        # no vote\n",
    "        if max_vote == 0:\n",
    "            res = [None, None, None, None]\n",
    "        else:\n",
    "            # check count\n",
    "            # if draw vote, check proba\n",
    "            if vote_list.count(max_vote) > 1:\n",
    "                chosen_cate = vote_proba_mean.index(max(vote_proba_mean))\n",
    "\n",
    "            # if not draw\n",
    "            else:\n",
    "                chosen_cate = vote_count.index(max_vote)\n",
    "\n",
    "            # append to result DataFrame\n",
    "            res = [topic_dict[chosen_cate], vote_proba_mean[chosen_cate], vote_list[chosen_cate], vote_probas[chosen_cate]]\n",
    "        new_row = pd.Series(res, index=column_names)\n",
    "        df = df.append(new_row, ignore_index=True)\n",
    "        \n",
    "    return df\n",
    "\n",
    "IRRE_ENCODE_1 = 0\n",
    "IRRE_ENCODE_2 = topic_encode['IRRELEVANT']\n",
    "\n",
    "def cate_vote_2(model_list, sample_size, truth):            \n",
    "    column_names = ['label', 'predict', 'mean_proba', 'models', 'probas']\n",
    "    final_df = pd.DataFrame(columns = column_names)\n",
    "    \n",
    "    all_votes = [[] for _ in range(sample_size)]\n",
    "    for clf in model_list:\n",
    "        print(clf.name)\n",
    "        if clf.is_preprocess is True:\n",
    "            test_x = dict_data[clf.test_set]\n",
    "        else:\n",
    "            test_x = dict_data['raw'][clf.test_set]\n",
    "        y_pred = clf.predict(test_x)\n",
    "        y_pred_proba = clf.predict_proba(test_x)\n",
    "        print(y_pred)\n",
    "        for i in range(sample_size):\n",
    "            cate, proba = y_pred[i], y_pred_proba[i]\n",
    "            clf_name = clf.name.split('.')[0]\n",
    "                            \n",
    "            # if level 1 and IRRELEVANT, add to vote\n",
    "            if clf.level ==1 and cate == IRRE_ENCODE_1:\n",
    "                all_votes[i].append(Vote(clf_name, IRRE_ENCODE_2, proba.max()))\n",
    "            elif clf.level == 2 and topic_decode[cate] in clf.topic:\n",
    "                all_votes[i].append(Vote(clf_name, cate, proba.max()))\n",
    "    vote_res = vote_proba_2(all_votes, topic_decode)\n",
    "    for i in range(sample_size):\n",
    "        final_res = [truth[i]] + vote_res.iloc[i].tolist()\n",
    "        new_row = pd.Series(final_res, index=column_names)\n",
    "        final_df = final_df.append(new_row, ignore_index=True)\n",
    "\n",
    "    return final_df\n",
    "\n",
    "res_df = cate_vote_2(model_list_1 + model_list_2, SAMPLE_SIZE, df['topic'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 137,
   "metadata": {},
   "outputs": [
    {
     "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>label</th>\n",
       "      <th>predict</th>\n",
       "      <th>mean_proba</th>\n",
       "      <th>models</th>\n",
       "      <th>probas</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.855460</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9415586552948044, 1.0, 0.6248227956728621]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.798582</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9294322124011538, 1.0, 0.46631374746266774]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>FOREX MARKETS</td>\n",
       "      <td>MONEY MARKETS</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>[clf_NB_2]</td>\n",
       "      <td>[0.9999999999997726]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.783383</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.8360490328877616, 1.0, 0.5140987736785823]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.984085</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9852943113674895, 1.0, 0.9669618390243637]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.834089</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9740126783506491, 1.0, 0.5282558161717601]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>496</th>\n",
       "      <td>SPORTS</td>\n",
       "      <td>SPORTS</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>[tfidf_logi_pipe, clf_NB_1, clf_NB_2, svm_prob...</td>\n",
       "      <td>[1.0, 1.0, 1.0, 0.9999999995165704, 0.99999995...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497</th>\n",
       "      <td>MONEY MARKETS</td>\n",
       "      <td>MONEY MARKETS</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>[clf_NB_2]</td>\n",
       "      <td>[1.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>498</th>\n",
       "      <td>SHARE LISTINGS</td>\n",
       "      <td>SHARE LISTINGS</td>\n",
       "      <td>0.972932</td>\n",
       "      <td>[tfidf_logi_pipe, clf_NB_1, clf_NB_2, svm_prob...</td>\n",
       "      <td>[0.8917285046808255, 1.0, 1.0, 0.9999995012957...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>499</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>MONEY MARKETS</td>\n",
       "      <td>0.997769</td>\n",
       "      <td>[clf_NB_2]</td>\n",
       "      <td>[0.9977686846157074]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>500 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "              label         predict  mean_proba  \\\n",
       "0        IRRELEVANT      IRRELEVANT    0.855460   \n",
       "1        IRRELEVANT      IRRELEVANT    0.798582   \n",
       "2     FOREX MARKETS   MONEY MARKETS    1.000000   \n",
       "3        IRRELEVANT      IRRELEVANT    0.783383   \n",
       "4        IRRELEVANT      IRRELEVANT    0.984085   \n",
       "..              ...             ...         ...   \n",
       "495      IRRELEVANT      IRRELEVANT    0.834089   \n",
       "496          SPORTS          SPORTS    1.000000   \n",
       "497   MONEY MARKETS   MONEY MARKETS    1.000000   \n",
       "498  SHARE LISTINGS  SHARE LISTINGS    0.972932   \n",
       "499      IRRELEVANT   MONEY MARKETS    0.997769   \n",
       "\n",
       "                                                models  \\\n",
       "0            [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "1            [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "2                                           [clf_NB_2]   \n",
       "3            [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "4            [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "..                                                 ...   \n",
       "495          [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "496  [tfidf_logi_pipe, clf_NB_1, clf_NB_2, svm_prob...   \n",
       "497                                         [clf_NB_2]   \n",
       "498  [tfidf_logi_pipe, clf_NB_1, clf_NB_2, svm_prob...   \n",
       "499                                         [clf_NB_2]   \n",
       "\n",
       "                                                probas  \n",
       "0        [0.9415586552948044, 1.0, 0.6248227956728621]  \n",
       "1       [0.9294322124011538, 1.0, 0.46631374746266774]  \n",
       "2                                 [0.9999999999997726]  \n",
       "3        [0.8360490328877616, 1.0, 0.5140987736785823]  \n",
       "4        [0.9852943113674895, 1.0, 0.9669618390243637]  \n",
       "..                                                 ...  \n",
       "495      [0.9740126783506491, 1.0, 0.5282558161717601]  \n",
       "496  [1.0, 1.0, 1.0, 0.9999999995165704, 0.99999995...  \n",
       "497                                              [1.0]  \n",
       "498  [0.8917285046808255, 1.0, 1.0, 0.9999995012957...  \n",
       "499                               [0.9977686846157074]  \n",
       "\n",
       "[500 rows x 5 columns]"
      ]
     },
     "execution_count": 137,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "res_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classification Report:\n",
      "                                  precision    recall  f1-score   support\n",
      "\n",
      "      ARTS CULTURE ENTERTAINMENT       0.38      1.00      0.55         3\n",
      "BIOGRAPHIES PERSONALITIES PEOPLE       0.52      0.73      0.61        15\n",
      "                         DEFENCE       0.63      0.92      0.75        13\n",
      "                DOMESTIC MARKETS       0.15      1.00      0.27         2\n",
      "                   FOREX MARKETS       0.57      0.79      0.66        48\n",
      "                          HEALTH       0.73      0.79      0.76        14\n",
      "                      IRRELEVANT       0.96      0.73      0.83       266\n",
      "                   MONEY MARKETS       0.55      0.67      0.61        69\n",
      "          SCIENCE AND TECHNOLOGY       0.50      0.33      0.40         3\n",
      "                  SHARE LISTINGS       0.56      0.71      0.63         7\n",
      "                          SPORTS       0.95      0.95      0.95        60\n",
      "\n",
      "                        accuracy                           0.76       500\n",
      "                       macro avg       0.59      0.78      0.64       500\n",
      "                    weighted avg       0.82      0.76      0.78       500\n",
      "\n",
      "[[  3   0   0   0   0   0   0   0   0   0   0]\n",
      " [  3  11   0   0   0   0   1   0   0   0   0]\n",
      " [  0   0  12   0   0   0   1   0   0   0   0]\n",
      " [  0   0   0   2   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0  38   0   2   7   0   1   0]\n",
      " [  0   0   0   1   0  11   1   0   1   0   0]\n",
      " [  0   8   7  10   7   3 195  30   0   3   3]\n",
      " [  0   0   0   0  22   0   1  46   0   0   0]\n",
      " [  1   0   0   0   0   1   0   0   1   0   0]\n",
      " [  0   0   0   0   0   0   2   0   0   5   0]\n",
      " [  1   2   0   0   0   0   0   0   0   0  57]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 936x504 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "DEFENCE accuarcy is:\t\t\t\t 0.9230769230769231\n",
      "DOMESTIC MARKETS accuarcy is:\t\t\t\t 1.0\n",
      "MONEY MARKETS accuarcy is:\t\t\t\t 0.6666666666666666\n",
      "FOREX MARKETS accuarcy is:\t\t\t\t 0.7916666666666666\n",
      "SPORTS accuarcy is:\t\t\t\t 0.95\n",
      "SHARE LISTINGS accuarcy is:\t\t\t\t 0.7142857142857143\n",
      "BIOGRAPHIES PERSONALITIES PEOPLE accuarcy is:\t\t\t\t 0.7333333333333333\n",
      "HEALTH accuarcy is:\t\t\t\t 0.7857142857142857\n",
      "ARTS CULTURE ENTERTAINMENT accuarcy is:\t\t\t\t 1.0\n",
      "SCIENCE AND TECHNOLOGY accuarcy is:\t\t\t\t 0.3333333333333333\n",
      "IRRELEVANT accuarcy is:\t\t\t\t 0.7330827067669173\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "report(res_df['label'], res_df['predict'])\n",
    "evaluate(res_df, 'label', 'predict')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "# recommandation\n",
    "with open('recommandation.txt', 'w') as f:\n",
    "    f.write('1 level_1 no weight\\n')\n",
    "    for topic in topic_list:\n",
    "        if topic != \"irrelavant\":\n",
    "            df_filter = res_df[res_df['predict'] == topic]\n",
    "            f.write(topic)\n",
    "            f.write('\\n')\n",
    "            f.write(str(df_filter.sort_values(by=\"mean_proba\")['label'].head(10)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [],
   "source": [
    "# recommandation\n",
    "with open('recommandation_1.txt', 'w') as f:\n",
    "    f.write('1 level_1 add_weight IRRE 3\\n')\n",
    "    for topic in topic_list:\n",
    "        if topic != \"irrelavant\":\n",
    "            df_filter = res_df[res_df['predict'] == topic]\n",
    "            f.write(topic)\n",
    "            f.write('\\n')\n",
    "            f.write(str(df_filter.sort_values(by=\"mean_proba\")['label'].head(10)))\n",
    "            f.write('\\n')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [],
   "source": [
    "# recommandation\n",
    "with open('recommandation_2.txt', 'w') as f:\n",
    "    f.write('1 level_1 add_weight IRRE 3 BIO 2 HEALTH 1 MONEY 3\\n')\n",
    "    for topic in topic_list:\n",
    "        if topic != \"irrelavant\":\n",
    "            df_filter = res_df[res_df['predict'] == topic]\n",
    "            f.write(topic)\n",
    "            f.write('\\n')\n",
    "            f.write(str(df_filter.sort_values(by=\"mean_proba\")['label'].head(10)))\n",
    "            f.write('\\n')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [],
   "source": [
    "# recommandation\n",
    "with open('recommandation_3.txt', 'w') as f:\n",
    "    f.write('3 level_1 add_weight BIO 2 HEALTH 1 MONEY 3\\n')\n",
    "    for topic in topic_list:\n",
    "        if topic != \"irrelavant\":\n",
    "            df_filter = res_df[res_df['predict'] == topic]\n",
    "            f.write(topic)\n",
    "            f.write('\\n')\n",
    "            f.write(str(df_filter.sort_values(by=\"mean_proba\")['label'].head(10)))\n",
    "            f.write('\\n')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {},
   "outputs": [],
   "source": [
    "# recommandation\n",
    "with open('recommandation_4.txt', 'w') as f:\n",
    "    f.write('3 level_1 add_weight all\\n')\n",
    "    for topic in topic_list:\n",
    "        if topic != \"irrelavant\":\n",
    "            df_filter = res_df[res_df['predict'] == topic]\n",
    "            f.write(topic)\n",
    "            f.write('\\n')\n",
    "            f.write(str(df_filter.sort_values(by=\"mean_proba\")['label'].head(10)))\n",
    "            f.write('\\n')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "svm_ir_clf.joblib\n",
      "[0 0 1 0 0 1 0 1 1 0 0 0 1 1 0 1 0 1 0 1 0 0 0 0 1 1 0 0 1 1 1 0 0 1 0 1 0\n",
      " 0 1 0 1 1 0 0 0 0 0 1 0 1 1 0 0 1 1 0 0 0 1 1 0 0 0 1 1 0 0 1 1 1 1 1 1 1\n",
      " 1 1 1 0 0 1 0 1 1 1 0 1 1 1 0 0 0 0 0 1 0 1 1 0 1 1 1 1 0 1 0 0 0 1 1 1 0\n",
      " 0 0 0 0 1 1 1 0 1 1 0 0 0 1 0 1 0 0 1 0 1 0 1 0 0 0 1 0 1 1 0 0 0 1 0 1 0\n",
      " 0 1 0 0 1 0 0 1 1 0 1 0 1 1 1 0 1 1 1 1 0 1 1 1 0 0 0 1 1 1 0 1 0 1 1 0 1\n",
      " 0 0 0 0 0 1 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 1 1 0 0 1 0 0 0 0 0 0 1 0 1 1 0\n",
      " 1 0 1 0 0 0 1 0 0 0 1 1 0 1 1 1 1 1 0 0 1 0 0 1 0 1 0 0 1 1 1 1 1 0 1 0 0\n",
      " 1 1 0 1 0 1 1 0 0 1 1 0 1 1 1 0 1 0 0 0 0 0 1 0 0 0 1 1 0 1 0 1 0 0 0 0 0\n",
      " 0 1 0 1 1 0 0 0 1 0 1 1 1 1 1 0 1 0 0 1 1 1 0 1 1 1 1 1 0 1 0 1 0 1 0 1 0\n",
      " 1 1 1 1 0 0 1 0 1 0 0 1 0 0 1 1 0 1 1 1 1 1 0 1 1 0 1 0 1 1 0 0 1 0 0 1 0\n",
      " 1 0 1 0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 1 0 0 1 1 1 0 1 0 1 0 0 1 1 0 1 0 0 0\n",
      " 0 0 0 1 0 0 0 0 1 0 1 1 1 0 1 0 1 0 1 0 0 1 0 1 0 1 0 0 0 1 0 1 1 0 1 1 1\n",
      " 0 0 1 0 0 0 0 1 0 0 0 0 0 0 1 0 1 0 1 1 1 0 1 0 1 0 0 1 0 0 1 0 1 0 1 0 1\n",
      " 1 0 1 1 1 1 1 0 1 0 1 0 1 1 0 1 1 0 0]\n",
      "rel_clf_tfidf_NB.joblib\n",
      "[0 0 1 0 0 1 0 1 1 1 0 0 1 1 0 1 0 0 0 1 0 1 0 0 1 1 0 1 1 1 1 1 0 1 0 1 0\n",
      " 0 1 0 1 1 0 0 0 0 1 1 0 1 1 0 0 1 1 0 0 0 0 1 1 0 0 1 1 0 0 1 1 1 0 1 1 1\n",
      " 1 1 1 0 0 1 0 1 1 1 0 1 0 1 1 0 0 0 0 1 0 1 1 0 1 1 0 1 0 1 0 0 1 1 0 1 0\n",
      " 0 0 0 0 1 1 1 0 1 1 0 0 0 1 0 1 0 0 1 0 1 0 1 0 0 0 1 0 0 1 0 0 1 1 0 1 0\n",
      " 0 1 0 0 1 0 1 1 1 0 1 0 1 1 1 0 1 0 0 1 0 0 1 1 0 0 0 1 1 1 0 1 0 1 1 0 0\n",
      " 0 0 0 0 0 1 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 1 1 0 0 1 0 0 0 1 0 0 1 0 1 0 1\n",
      " 1 0 1 0 1 0 1 0 0 0 1 1 0 1 1 1 0 1 0 0 1 0 0 1 0 1 0 0 1 1 1 1 1 0 1 1 0\n",
      " 1 1 0 1 0 1 1 0 0 1 0 0 1 1 1 0 1 0 0 0 0 0 1 1 0 0 1 1 0 1 0 1 0 0 0 0 0\n",
      " 0 1 0 1 1 0 0 0 0 0 1 0 1 0 1 0 1 0 0 1 1 1 0 1 1 1 1 1 0 1 0 1 1 1 0 1 1\n",
      " 0 1 1 1 0 0 1 0 1 0 0 1 0 0 1 1 0 1 1 1 1 1 0 1 1 0 1 0 1 1 0 0 1 0 0 1 0\n",
      " 1 0 1 0 1 0 0 1 0 0 0 0 1 0 1 1 0 0 1 0 0 0 1 1 0 1 0 1 0 1 1 1 0 0 0 0 0\n",
      " 0 0 0 1 0 0 0 0 1 0 0 1 1 0 1 0 1 0 1 0 0 1 1 1 0 1 0 0 0 0 1 1 1 0 1 1 1\n",
      " 0 0 1 0 0 0 0 1 0 0 0 1 0 0 1 0 1 0 1 1 1 0 1 0 0 0 0 0 0 1 1 0 1 1 1 0 1\n",
      " 1 0 1 1 1 0 1 0 1 0 1 0 0 1 0 1 1 1 1]\n",
      "best_logi.joblib\n",
      "[0 0 1 0 0 1 0 1 1 0 0 0 1 1 0 1 0 1 0 1 0 0 0 0 1 1 0 0 1 1 1 0 1 1 0 1 0\n",
      " 0 1 0 1 1 0 0 0 0 1 1 0 1 1 0 1 1 1 0 0 0 1 1 0 0 0 0 1 0 0 1 1 1 1 1 1 1\n",
      " 0 1 1 0 0 1 0 1 1 1 0 1 1 1 0 0 0 0 0 1 0 1 1 0 1 1 1 1 0 1 0 0 1 1 1 1 0\n",
      " 0 0 0 0 1 1 1 0 1 1 0 0 0 1 0 1 0 0 1 0 1 0 1 0 0 0 1 0 0 1 0 0 0 1 0 1 0\n",
      " 0 1 0 0 1 0 1 1 1 0 1 0 1 1 1 0 1 1 1 1 0 1 1 1 0 0 0 1 1 1 0 1 0 1 1 0 0\n",
      " 0 0 0 0 0 1 0 1 1 1 1 0 0 0 0 1 1 1 1 0 1 1 1 0 0 1 0 0 0 1 0 0 1 0 1 1 0\n",
      " 1 0 1 0 0 0 1 0 0 0 1 1 1 1 1 1 0 1 0 0 1 0 0 0 0 1 1 0 1 1 1 1 1 0 1 1 1\n",
      " 1 1 0 1 0 1 1 0 0 1 1 0 1 1 1 0 1 0 0 0 0 0 1 0 0 0 1 1 0 1 0 1 0 0 0 0 1\n",
      " 1 1 0 1 1 0 0 0 0 1 1 1 1 0 1 0 1 0 0 1 1 1 0 1 1 1 1 1 0 1 0 1 0 1 0 1 1\n",
      " 1 1 1 1 0 0 1 0 0 0 0 1 0 0 1 1 0 1 1 1 1 1 0 1 1 0 1 0 1 1 0 0 1 0 0 1 0\n",
      " 1 0 1 0 1 0 1 1 0 0 0 0 1 0 0 1 0 0 1 0 0 1 1 1 0 1 0 1 0 0 1 1 0 0 0 0 0\n",
      " 0 0 1 1 0 0 0 0 1 0 0 1 1 0 1 0 1 0 1 0 0 1 1 1 0 1 0 0 0 1 1 1 1 0 1 0 1\n",
      " 0 0 1 0 0 0 0 1 0 0 0 0 0 0 1 0 1 0 1 1 1 0 1 0 0 0 0 1 0 0 1 0 1 1 1 0 1\n",
      " 1 0 1 1 1 1 1 1 1 0 1 0 1 1 0 1 1 1 0]\n",
      "tfidf_logi_pipe.joblib\n",
      "[ 6  6  7  3  6  4  2 10  7  6  6  6 10  6  3  7  6  9  3 10  4  7  6  6\n",
      "  7  1  6  6  4  4  4  7  7  7  6 10  3  2  4  0 10  7  6  6  3  6  7  4\n",
      "  6  4  4  3  7  4  7  0  6  6  2  4  7  9  6  7  4  6  4 10 10  4  7  4\n",
      " 10 10  1  2  4  6  2 10  9  6  7  4  6  7  7  4  7  6  6  6  4  7  6 10\n",
      " 10  6  7  7  9  7  6  0  9  7  2 10  5 10  9  8  4  9  6  2  8  7  6 10\n",
      "  8  6  7  6  4  7  7  2  5 10  3  4  6  4  6  6  6  7  6  3  7  6  3  6\n",
      "  3  6  7  6  6  4  6  9 10  6  9 10 10  6  4  1  5 10 10  6  7  9  9  9\n",
      "  6  6  4  4  6  6  6  7  7  4  6  7  3  4  7  9  7  6  6  9  6  6  7  6\n",
      "  4 10 10  4  3  6  6  6 10  4  0  4  9  2  7  7  7  7  4  9  2  7  9  6\n",
      "  6  4  6  0  2  6  7  6  4  6  9  6  7  6  2  6  4  4  5 10  7 10  2  4\n",
      "  4  6  4  6  6  7  9  4  7  3  7 10  7 10  7  6  0  5  6 10  7  3 10  1\n",
      "  7  7  3  1  7  2  6  4  2 10  6  4  6  6  7  3  2 10  1  7  6  4 10  8\n",
      "  0  1 10  7  6  6  6  3  1  4  6 10 10  6  7  6  7  7  5  6 10  5 10  6\n",
      " 10  6  1  7 10 10  6  7 10  7  4  7  6  4  6  7  6  0  4 10  3  0  7 10\n",
      "  4  6  9  7  6  3  6  9  4  6  6 10 10  6  4  4  7  5  4  7 10 10  6  7\n",
      "  6 10  7  9  6  4  9  7  4  6  7  6  5  6  4  9  7  1  9  6  4  2  4  6\n",
      "  7 10  5  6  4  6  6  4  4  4  2  1  7  7  9  0  7  4  6  2  6  6  3  6\n",
      "  7  3  5  6  6  6  7  7  6  6 10 10  6 10  3  4  9  5  0  6  8  7 10  6\n",
      "  0  4  6  7  5  7  7  1  6 10  4  7  6  9  5  1  9  6  9  0  9  3  2  6\n",
      "  6  6  7  6  4  6  4 10  4  6  7  7  7  6  7  9  6  1  4  6  4  6 10  6\n",
      " 10  8  1  4  7  4  2  1  3  7  9 10  6  3  4  8 10  7  9  6]\n",
      "clf_NB_1.joblib\n",
      "[ 3  2  4  3  4  4  2 10  7  4  4  3 10 10  3  7  2  9  3 10  4  4  3  9\n",
      "  7  1  9  4  4  4  4  4  7  7  7 10  7  3  4  9 10  4  9  7  2  2  7  4\n",
      "  3  7  4  3  7  4  4  9  9  9  2  4  4  9  9  4  4  3  4 10 10  4  7  4\n",
      " 10 10  1  2  4  3  2 10  9  1  7  4  7  7  4  4  4  3  3  4  3  7  6 10\n",
      " 10  9  4  7  9  7  5  0  9  7  2 10  5 10  4  5  4  4  4  2  8  7  2 10\n",
      "  5  9  7  9  4  7  7  8  9 10  3  4  9  4  6  3  9  7  9  4  7  4  3  5\n",
      "  2  6  4  2  4  4  3  9 10  9  9 10 10  4  4  2  5 10 10  9  4  9  9  9\n",
      "  4  2  4  4  2  9  7  7  4  4  4  7  3  4  7  9  4  3  4  9  3  4  7  3\n",
      "  4 10 10  4  3  2  3  7 10  4  0  4  9  2  7  7  4  4  4  9  2  2  9  6\n",
      "  4  4  7  7  2  8  7  1  4  9  4  3  4  9  2  9  4  4  5 10  7 10  2  4\n",
      "  2  9  4  7  3  4  9  4  7  3  4 10  4 10  7  2  4  1  2 10  7  3 10  2\n",
      "  7  7  3  2  7  2  2  4  2 10  2  7  2  7  4  3  9 10  2  4  9  4 10  3\n",
      "  0  9 10  7  7  6  3  3  2  4  2 10 10  3  4  9  7  4  5  7 10  5  1  9\n",
      " 10  7  2  7 10 10  2  7 10  7  4  7  4  4  3  7  7  0  2 10  3  0  7 10\n",
      "  4  9  4  7  9  2  9  4  4  9  5 10 10  4  4  4  7  1  4  7 10 10  4  7\n",
      "  4 10  7  9  7  4  9  4  4  4  7  7  5  4  4  9  7  1  4  7  4  2  4  9\n",
      "  4 10  5  4  4  2  6  4  4  4  3  1  7  4  2  9  7  4  3  2  2  7  3  3\n",
      "  3  3  5  9  4  3  4  4  9  9 10 10  9 10  3  7  2  5  2  4  5  7 10  9\n",
      "  0  4  2  7  5  4  7  1  7 10  4  4  9  9  5  4  9  9  9  0  7  3  2  1\n",
      "  4  3  4  3  4  2  4 10  4  9  7  7  7  3  7  9  9  1  4  9  4  5 10  3\n",
      " 10  8  1  4  7  4  2  1  3  7  9 10  9  3  7  8 10  7  9  4]\n",
      "clf_NB_2.joblib\n",
      "[ 9  6  7  6  6  7  6 10  7  7  9  6 10 10  7  7  6  9  3 10  7  7  7  6\n",
      "  7  1  9  7  7  7  7  4  7  7  7 10  7  6  7  9 10  4  6  7  9  8  7  7\n",
      "  3  7  7  3  7  7  7  9  9  7  2  7  4  9  9  4  7  3  7 10 10  7  7  4\n",
      " 10 10  5  2  4  3  2 10  9  1  7  4  7  7  7  7  7  3  3  6  3  7  6 10\n",
      " 10  7  7  7  9  7  9  0  9  7  2 10  5 10  9  5  7  9  6  2  5  7  6 10\n",
      "  5  3  9  9  7  7  7  6  9 10  3  4  3  7  6  9  9  7  9  3  7  4  3  9\n",
      "  3  6  4  2  9  7  3  6 10  9  9 10 10  6  7  2  5 10 10  9  7  9  9  9\n",
      "  7  2  7  7  2  9  7  7  7  4  4  7  3  7  7  9  7  9  6  9  4  6  7  6\n",
      "  4 10 10  4  7  6  3  7 10  7  1  4  9  2  7  7  7  7  4  9  2  6  9  7\n",
      "  7  7  7  7  4  8  7  2  7  6  7  3  7  9  2  9  7  4  5 10  7 10  2  7\n",
      "  2  9  4  7  7  7  6  4  7  3  4 10  7 10  7  3 10  6  2 10  7  3 10  2\n",
      "  7  7  3  2  7  2  9  7  2 10  5  7  2  7  7  3  9 10  2  7  9  7 10  3\n",
      "  0  3 10  6  6  6  3  3  2  7  2 10 10  3  7  9  7  4  5  7 10  5  8  9\n",
      " 10  9  2  7 10 10  5  7 10  7  4  7  4  7  9  7  7  0  6 10  3  9  7 10\n",
      "  4  9  9  7  9  2  9  9  4  9  9 10 10  7  7  7  7  5  7  7 10 10  9  7\n",
      "  4  6  7  9  7  7  9  4  4  7  7  9  5  7  4  9  7  1  9  6  9  2  7  9\n",
      "  7 10  5  4  4  6  3  6  4  7  6  1  7  7  2  6  7  4  3  2  2  7  3  9\n",
      "  3  3  5  9  7  6  7  7  9  6 10 10  9 10  3  7  9  5  1  9  5  7 10  9\n",
      "  1  7  2  9  5  7  7  1  7 10  4  7  7  9  5  6  9  9  9  0  7  3  2  6\n",
      "  7  6  7  6  7  2  4 10  4  9  7  7  7  3  7  9  9  5  7  9  7  5 10  3\n",
      " 10  5  1  7  7  7  2  5  3  7  9 10  7  3  7  9 10  7  9  7]\n",
      "clf_NB_3.joblib\n",
      "[ 3  0  7  6  6  4  6 10  7  4  4  6 10 10  3  7  6  9  3 10  6  4  7  6\n",
      "  7  1  6  4  7  4  4  4  1  7  7 10  7  6  4  6 10  4  6  7  2  2  7  4\n",
      "  3  7  4  3  7  4  7  6  9  9  2  7  4  9  6  4  4  3  4 10 10  4  7  4\n",
      " 10 10  1  2  4  3  2 10  9  1  7  4  7  7  7  4  4  3  3  6  3  7  6 10\n",
      " 10  7  4  7  9  7  0  0  9  7  2 10  5 10  4  5  4  4  6  2  8  7  6 10\n",
      "  5  9  6  9  4  7  7  2  6 10  3  4  6  4  6  6  9  7  6  3  7  4  3  5\n",
      "  1  6  4  2  4  4  3  6 10  9  9 10 10  6  4  2  5 10 10  6  4  9  9  9\n",
      "  4  6  4  4  2  9  7  7  4  4  2  7  3  4  7  9  4  6  4  9  3  6  7  3\n",
      "  4 10 10  4  6  1  6  9 10  4  0  4  6  2  7  7  6  4  4  9  2  2  9  6\n",
      "  4  4  6  7  2  8  7  1  4  9  4  6  4  6  2  9  4  4  5 10  7 10  2  4\n",
      "  2  9  4  7  6  7  6  4  7  3  4 10  4 10  7  6 10  1  2 10  7  3 10  5\n",
      "  7  7  3  1  7  2  2  4  2 10  6  7  2  7  6  3  9 10  1  7  6  4 10  6\n",
      "  0  6 10  7  6  6  3  3  1  4  1 10 10  3  6  9  7  6  5  7 10  5  1  9\n",
      " 10  6  2  7 10 10  2  7 10  7  4  7  4  4  6  7  7  0  6 10  3  0  7 10\n",
      "  4  6  4  7  6  2  6  6  4  9  6 10 10  7  4  4  7  1  4  7 10 10  6  7\n",
      "  4 10  7  9  6  4  9  6  4  7  7  6  5  4  4  9  7  1  4  6  4  2  4  6\n",
      "  7 10  5  6  4  6  6  6  4  4  6  1  7  7  2  6  7  4  6  2  2  7  3  6\n",
      "  7  3  5  9  4  6  4  4  6  6 10 10  9 10  3  7  6  5  0  6  5  7 10  9\n",
      "  0  4  2  6  5  4  7  1  9 10  4  7  6  9  5  6  9  9  9  0  7  3  2  1\n",
      "  4  6  4  6  4  2  4 10  4  9  7  7  7  3  7  9  9  1  4  6  4  5 10  3\n",
      " 10  5  1  4  7  4  2  1  3  7  6 10  6  3  7  6 10  7  9  4]\n",
      "svm_proba_1.joblib\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 3  6  4  6  6  4  2 10  7  9  9  6 10 10  6  7  6  9  3 10  6  4  7  6\n",
      "  7  1  6  7  4  4  4  4  1  7  6 10  3  6  4  6 10  4  6  8  2  1  7  4\n",
      "  3  4  4  3  7  4  4  6  9  6  2  7  4  9  9  4  4  6  4 10 10  4  7  4\n",
      " 10 10  1  2  4  6  2 10  9  1  7  4  6  7  7  4  4  6  3  6  6  4  6 10\n",
      " 10  6  4  7  9  7  6  1  9  7  2 10  5 10  4  6  4  9  6  2  8  7  2 10\n",
      "  5  9  7  6  4  4  4  6  6 10  6  4  6  4  6  6  6  7  6  3  7  6  3  6\n",
      "  1  6  4  2  4  4  3  9 10  9  9 10 10  9  4  6  5 10 10  6  4  9  9  6\n",
      "  1  2  4  7  2  6  7  4  4  4  2  7  6  4  7  9  4  6  6  9  3  6  7  3\n",
      "  4 10 10  4  3  1  6  9 10  4  0  4  6  2  7  4  4  4  4  9  2  2  9  6\n",
      "  4  4  6  7  2  8  7  1  4  6  9  6  4  9  2  9  4  4  5 10  7 10  2  4\n",
      "  2  9  4  7  6  7  9  4  7  3  4 10  3 10  7  6 10  1  2 10  7  6 10  1\n",
      "  7  7  3  1  7  2  2  4  2 10  5  4  2  7  4  3  9 10  2  4  6  4 10  6\n",
      "  0  6 10  7  6  6  3  3  1  4  1 10 10  3  7  9  7  4  5  6 10  5 10  9\n",
      " 10  6  2  7 10 10  6  7 10  7  4  7  4  4  6  7  7  0  6 10  3  0  7 10\n",
      "  4  6  9  7  6  2  6  9  4  6  6 10 10  9  4  4  7  1  4  7 10 10  6  7\n",
      "  4 10  7  9  6  4  9  6  4  6  7  6  5  4  4  9  7  1  9  6  4  2  4  9\n",
      "  4 10  5  9  4  2  6  2  4  4  9  1  7  7  2  6  7  4  3  2  2  8  3  6\n",
      "  6  3  5  9  4  6  7  7  6  6 10 10  9 10  3  4  6  5  2  6  5  7 10  6\n",
      "  0  4  2  7  5  7  7  1  9 10  4  4  9  9  5  1  9  9  9  0  9  3  2  1\n",
      "  6  3  4  6  4  6  4 10  4  9  7  7  4  6  7  9  6  6  4  6  4  5 10  6\n",
      " 10  5  1  4  7  4  2  1  3  7  6 10  6  3  7  6 10  7  9  4]\n",
      "svm_proba_2.joblib\n",
      "[ 6  6  4  6  6  4  6 10  7  6  9  6 10 10  6  7  6  9  3 10  6  4  6  6\n",
      "  7  0  6  6  4  4  4  4  6  7  6 10  6  6  4  6 10  6  6  6  6  6  7  4\n",
      "  6  7  4  6  7  4  4  6  6  6  2  7  4  9  6  4  4  6  6 10 10  4  6  4\n",
      " 10 10  1  2  4  6  6 10  6  1  7  4  6  7  7  4  6  6  6  6  6  7  6 10\n",
      " 10  6  4  7  9  7  6  6  6  7  2 10  5 10  6  6  6  6  6  2  8  7  6 10\n",
      "  5  6  6  6  4  6  4  6  6 10  6  4  6  4  6  6  6  7  6  3  4  6  6  6\n",
      "  1  6  4  6  6  4  6  6 10  6  6 10 10  6  4  6  5 10 10  6  4  9  6  6\n",
      "  1  2  4  7  6  6  6  4  4  4  6  7  6  4  4  6  4  6  6  6  6  6  7  3\n",
      "  4 10 10  4  6  6  6  6 10  4  0  4  6  2  7  4  6  7  4  9  2  6  9  6\n",
      "  6  4  6  7  6  8  7  1  4  6  6  6  4  6  2  6  4  4  5 10  7 10  2  4\n",
      "  2  6  4  6  6  7  6  4  6  3  4 10  4 10  7  6 10  1  2 10  7  6 10  6\n",
      "  7  7  3  1  7  2  6  4  2 10  6  4  6  6  6  3  6 10  6  6  6  4 10  6\n",
      "  0  6 10  6  6  6  6  3  1  4  6 10 10  6  6  6  7  6  5  6 10  5 10  6\n",
      " 10  6  2  7 10 10  6  7 10  7  4  7  6  4  6  7  7  0  6 10  5  6  7 10\n",
      "  4  6  6  7  6  2  6  6  4  6  6 10 10  6  4  4  7  1  4  6 10 10  6  7\n",
      "  6 10  7  9  6  4  9  6  4  6  7  6  5  6  4  6  7  1  6  6  4  2  6  6\n",
      "  4 10  5  6  4  2  6  2  6  4  6  1  6  7  6  6  4  4  6  6  6  6  3  6\n",
      "  6  3  5  6  6  6  6  7  6  6 10 10  6 10  3  4  6  5  6  6  5  7 10  6\n",
      "  0  6  6  6  5  7  7  1  6 10  4  4  6  9  5  6  6  6  9  0  6  3  2  6\n",
      "  6  3  4  6  4  6  4 10  4  6  7  7  6  6  6  9  6  6  4  6  4  5 10  6\n",
      " 10  5  1  4  7  4  2  1  3  7  6 10  6  3  7  6 10  7  6  6]\n",
      "1_article_words.model\n",
      "[ 6  6  7  6  6  4  7 10  7  6  6  6 10 10  3  6  6  9  3 10  4  6  6  6\n",
      "  4  6  6  9  6  7  7  6  6  7  6 10  3  6  7  9  6  4  6  6  6  6  7  4\n",
      "  6  4  7  6  6  7  4  6  6  6  0  7  6  6  6  9  7  6  6 10 10  7  7  6\n",
      " 10 10  6  2  4  6  6 10  9 10  7  7  6  7  7  4  4  6  6  0  6  7  6 10\n",
      " 10  6  7  7  9  7  6  6  6  6 10 10  6 10  4  6  6  9  6  6  6  7  4 10\n",
      "  5  6  1  6  4  6  4  6  6 10  6  7  7  7  6  6  6  7  6  6  4  6  6  6\n",
      "  6  6  7  6  9  4  6  6 10  6  9 10 10  9  4  6  5 10 10  6  4  9  9  4\n",
      "  6  2  7  7  6  6  6  7  4  4  2  7  3  4  4  6  6  6  6  6  6  6  7  6\n",
      "  4 10 10  9  6  8  6  6 10  7  0  4  6  2  7  4  6  6  4  6  2  1  9  6\n",
      "  6  7  6  6  6  6  6  6  7  6  6  6  4  6  2  6  7  7  5 10  7 10  2  4\n",
      "  6  6  4  7  6  8  9  7  6  6  7  6  4 10 10  6  6  0  6 10  7  6 10  6\n",
      "  6  7  6  6  7  5  6  4  6 10  6  7  6  6  6  6  6 10  1 10  6  4 10  6\n",
      "  6  6 10  6  6  6  6  6 10  4  3 10 10  6  7  6  6  4  3  6 10  1 10  7\n",
      " 10  6  6  7 10  6  0  7 10  7  6  4  6  4  6  7  4  0  6 10  6  6  7 10\n",
      "  4  6  6  7  6  2  6  6  7  6  6 10 10  6  7  7  7  5  7  6 10 10  6  7\n",
      "  6 10  7  6  6  4  6  6  6  6  7  6  5  6  4  6  7  6  7  6  4  2  6  6\n",
      "  6 10  5  6  4  2  6  6  7  4  6 10  7  7  6  6  6  4  6  1  0  6  6  6\n",
      "  6  2  5  6  6  6  6  7  6  6 10 10  6 10  6  4  6  1  6  6  0  6 10  6\n",
      "  0  6  6  6  5  6  7  6  6 10  7  4  6  6  6  6  4  6  6  6  6  6  6  6\n",
      "  6  6  7  6  7  6  7 10  7  6  7  6  4  6  1  9  6  6  4  6  4  6 10  6\n",
      " 10  8 10  7  6  4  2  1  6  7  6 10  6  6  7  2 10 10  6  6]\n",
      "1_key_word_100.model\n",
      "[ 6  6  7  6  6  4  6 10  7  6  6  6 10 10  6  7 10  6  6 10  7  9  6  6\n",
      "  7  9  7  6  7  4  7  6  6  7  6 10  3  6  7  6 10  4  6  6  6  1  7  4\n",
      "  6  7  4  6  6  7  4  6  6  6  2  4  6  6  6  9  7  6  7 10 10  7  7  6\n",
      " 10 10  1  2  7  6  6 10  9 10  7  4  6  7  7  4  7  6  6  6  6  6  6 10\n",
      " 10  6  4  6  9  7  6  6  6  7  6 10  6 10  6  5  6  6  6  2  6  7  6 10\n",
      "  5  6  7  6  4  6  4  6  6 10  6  7  6  7  6  6  6  7  6  6  4  6  6  6\n",
      "  2  1  4  6  6  4  6  6 10  6  6 10 10  6  7  6  6 10 10  6  6  9  9  4\n",
      "  6  2  7  7  2  6  6  7  7  4  6  7  6  4  7  6  7  6  6  6  6  6  7  3\n",
      "  4 10 10  6  6  6  6  7 10  7  0  4  4  6  7  4  6  6  4  6  6  6  9  6\n",
      "  6  4  6  6  6  6  7  6  7  6  6  6  7  6  2  6  7  7  6 10  7 10  6  6\n",
      "  6  6  7  7  6  6  7  4  5  6  7 10  7 10 10  6  6  0  6 10  7  6 10  6\n",
      "  7  7  6  6  7  2  6  7  6 10  6  7  6  6  6  6  6 10  1 10  6  6 10  6\n",
      "  1  6 10  6  6  6  6  3 10  4  2 10 10  6  6  6  7  7  6  7 10  5  6  6\n",
      " 10  6  6  7 10  6  6  7  6  4  7  7  6  7  6  7  4  0  6 10  3  7  7 10\n",
      "  4  6  6  7  6  6  6  6  4  6  4 10  6  6  6  4  7  5  7  6 10 10  6  7\n",
      "  6 10  7  6  6  7  6  6  7  6  7  6  5  6  4  6  6  0  6  6  7  2  6  6\n",
      "  6 10  6  6  4  2  6  6  6  4  6  4  6  7  3  4  7  4  2  5  5  6  6  6\n",
      "  6  6 10  6  6  6  6  7  6  6 10  6  6  6  6  4  6  6  6  6  6  4 10  6\n",
      "  1  6  6  6  5  6  7  5  6 10  4  7  6  6  6  6  6  6  6  1  6  6  6  6\n",
      "  6  6  6 10  4  6  4 10  7  6  6  6  4  6  6  9  6  6  4  6  7 10 10  6\n",
      " 10  8  1  7  7  4  2  1  3  7  6 10  6  6  7  6 10 10  6  6]\n",
      "2_article_words.model\n",
      "[ 6  6  7  6  6  7  7 10  7  6  6  6 10 10  6  7  6  9  7 10  6  6  6  6\n",
      "  4  6  6  6  4  4  7  6  6  7  6 10  6  7  7  6 10  4  6  6  6  5  6  4\n",
      "  6  4  4  6  6  4  4  6  6  6  2  7  7  9  6  9  4  6  6 10  2  7  6  4\n",
      " 10 10  1  8  6  6  6 10  9 10  7  7 10  7  7  4  6  8  6  4  7  7  6 10\n",
      " 10  6  7  7  6  7  1  6  6  4  2 10  5 10 10  8  4  9  8  6  3  7  7 10\n",
      "  5  6  6  6  4  6  7  6  6 10 10  7  6  7  6  6  6  7  6  6  4  6  7  6\n",
      "  1  6  7  0  6  7  6  9  1  6  6 10 10  6  7  6  8 10 10  6  4  6  9  6\n",
      "  7  4  6  6  7  6  6  7  4  4  4  4  6  4  4  6  7  6  6  9  6  6  7  6\n",
      "  7 10 10  7  7  5  7  6 10  7  0  7  6  6  7  4  6  6  7  6  2  6  6  6\n",
      "  4  4  6  6  2  6  7  6  7  1  3  6  4  9  6  6  7  6  6  6  6 10 10  4\n",
      "  1  9  7  6  6  7  6  7  6  6  7 10  6 10  7  6  9  1  6 10  7  6 10  6\n",
      "  7  7  6  5  7  2  6  7  2 10  6  4  0  1  6  6  6 10  6  6  6  4 10  6\n",
      "  0  6 10  6  6  6  6  0 10  4  6 10  6  6  6  6  7  6  6  6 10  6  1  6\n",
      " 10  6  6  7 10  6  1  7 10  7  4  4  4  7  6  7  4  0  4 10  3  5  7  6\n",
      "  4  6  6  7  6  1  6  6  4  6  6 10 10  6  7  7  7  5  7  6 10 10  6  7\n",
      "  6  6  7  9  6  4  6  6  7  6  7  5  6  6  7  9  6  6  6  6  6  1  6  6\n",
      "  6 10  6  6  4  6  6  4  7  4  6  6  7  4 10  6  4  7  6  2  1  6  6  6\n",
      "  6  6  6  6  6  6  6  7  6  6 10 10  6 10  6  4  6  1  1 10  8  6 10  6\n",
      " 10  6  9  6  5  6  7  1  6 10  4  7  6  9  5  6  6  6  9  0  6  6  6  1\n",
      "  6  6  4 10  7  1  7 10  7  6  7  6  4  6  7  9  6  6  7  9  4 10 10  6\n",
      " 10  5  6  4  6  4  2 10  3  7  1 10  6  6  7  6 10  7  9  6]\n"
     ]
    }
   ],
   "source": [
    "# ADD_WEIGHT_LIST = [0 for _ in topic_list]\n",
    "# ADD_WEIGHT_LIST[topic_encode['BIOGRAPHIES PERSONALITIES PEOPLE']] = 2\n",
    "# ADD_WEIGHT_LIST[topic_encode['HEALTH']] = 1\n",
    "# ADD_WEIGHT_LIST[topic_encode['MONEY MARKETS']] = 3\n",
    "\n",
    "topic_count = {}\n",
    "for topic in topic_list:\n",
    "    topic_count[topic] = 0\n",
    "    \n",
    "for model in model_list_1 + model_list_2:\n",
    "    for topic in model.topic:\n",
    "        topic_count[topic] += 1\n",
    "# ADD_WEIGHT_LIST = [topic_count[max(topic_count)]-i for i in topic_count.values()]\n",
    "\n",
    "ADD_WEIGHT_LIST = [topic_count[max(topic_count)]/i for i in topic_count.values()]\n",
    "\n",
    "def vote_proba_3(votes, topic_dict, weights=None):\n",
    "    cate_count = len(topic_dict)\n",
    "\n",
    "    column_names = ['predict', 'mean_proba', 'models', 'probas']\n",
    "    df = pd.DataFrame(columns = column_names)\n",
    "\n",
    "    for num, sample in enumerate(votes): \n",
    "        # vote_list to count vote, init with 11 zeros\n",
    "        vote_list = [[] for _ in range(cate_count)]\n",
    "        vote_probas = [[] for _ in range(cate_count)]\n",
    "        for vote in sample:\n",
    "            vote_list[vote.cate].append(vote.model)\n",
    "            vote_probas[vote.cate].append(vote.proba)\n",
    "        vote_count = [len(cate_list) for cate_list in vote_list]\n",
    "        vote_proba_mean = [0 for _ in range(cate_count)]\n",
    "        for i in range(cate_count):\n",
    "            if len(vote_probas[i]) > 0:\n",
    "                vote_proba_mean[i] = sum(vote_probas[i])/len(vote_probas[i])\n",
    "                \n",
    "        ###### add weight ######\n",
    "#         if weights is not None:\n",
    "#             for i in range(len(vote_count)):\n",
    "#                 if vote_count[i] > 0:\n",
    "#                     vote_count[i] += weights[i]\n",
    "        if weights is not None:\n",
    "            for i in range(len(vote_count)):\n",
    "                vote_count[i] *= weights[i]\n",
    "        ########################\n",
    "            \n",
    "        # get category chosen by vote\n",
    "        max_vote = max(vote_count)\n",
    "        \n",
    "        # no vote\n",
    "        if max_vote == 0:\n",
    "            res = [None, None, None, None]\n",
    "        else:\n",
    "            # check count\n",
    "            # if draw vote, check proba\n",
    "            if vote_list.count(max_vote) > 1:\n",
    "                chosen_cate = vote_proba_mean.index(max(vote_proba_mean))\n",
    "\n",
    "            # if not draw\n",
    "            else:\n",
    "                chosen_cate = vote_count.index(max_vote)\n",
    "\n",
    "            # append to result DataFrame\n",
    "            res = [topic_dict[chosen_cate], vote_proba_mean[chosen_cate], vote_list[chosen_cate], vote_probas[chosen_cate]]\n",
    "        new_row = pd.Series(res, index=column_names)\n",
    "        df = df.append(new_row, ignore_index=True)\n",
    "        \n",
    "    return df\n",
    "\n",
    "def irre_vote_3(model_list, sample_size):\n",
    "    all_votes = [[] for _ in range(sample_size)]\n",
    "    for clf in model_list:\n",
    "            print(clf.name)\n",
    "            if clf.is_preprocess is True:\n",
    "                test_x = dict_data[clf.test_set]\n",
    "            else:\n",
    "                test_x = dict_data['raw'][clf.test_set]\n",
    "            y_pred = clf.predict(test_x)\n",
    "            y_pred_proba = clf.predict_proba(test_x)\n",
    "            print(y_pred)\n",
    "            for i in range(sample_size):\n",
    "                cate, proba = y_pred[i], y_pred_proba[i]\n",
    "\n",
    "                clf_name = clf.name.split('.')[0]\n",
    "                # create namedtuple (model_name, category, category_proba)\n",
    "                all_votes[i].append(Vote(clf_name, cate, proba.max()))\n",
    "    topic_dict = {0: 'IRRELEVANT', 1: 'RELEVANT'}\n",
    "    return vote_proba_3(all_votes, topic_dict, [1, 0])\n",
    "\n",
    "def cate_vote_3(model_list, sample_size, irre_res, truth):\n",
    "    \n",
    "    column_names = ['label', 'predict', 'mean_proba', 'models', 'probas']\n",
    "    final_df = pd.DataFrame(columns = column_names)\n",
    "    \n",
    "    all_votes = [[] for _ in range(sample_size)]\n",
    "    for clf in model_list:\n",
    "        print(clf.name)\n",
    "        if clf.is_preprocess is True:\n",
    "            test_x = dict_data[clf.test_set]\n",
    "        else:\n",
    "            test_x = dict_data['raw'][clf.test_set]\n",
    "        y_pred = clf.predict(test_x)\n",
    "        y_pred_proba = clf.predict_proba(test_x)\n",
    "        print(y_pred)\n",
    "        for i in range(sample_size):\n",
    "            cate, proba = y_pred[i], y_pred_proba[i]\n",
    "            if topic_decode[cate] in clf.topic:\n",
    "                            \n",
    "                clf_name = clf.name.split('.')[0]\n",
    "\n",
    "                all_votes[i].append(Vote(clf_name, cate, proba.max()))\n",
    "    rela_res = vote_proba_3(all_votes, topic_decode, ADD_WEIGHT_LIST)\n",
    "    for i in range(sample_size):\n",
    "        if irre_res['predict'].iloc[i] == 'IRRELEVANT':\n",
    "            final_res = [truth[i]] + irre_res.iloc[i].tolist()\n",
    "        else:\n",
    "            final_res = [truth[i]] + rela_res.iloc[i].tolist()\n",
    "        new_row = pd.Series(final_res, index=column_names)\n",
    "        final_df = final_df.append(new_row, ignore_index=True)\n",
    "\n",
    "    return final_df\n",
    "\n",
    "irre_df = irre_vote_3(model_list_1, SAMPLE_SIZE)\n",
    "res_df = cate_vote_3(model_list_2, SAMPLE_SIZE, irre_df, df['topic'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {},
   "outputs": [
    {
     "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>predict</th>\n",
       "      <th>mean_proba</th>\n",
       "      <th>models</th>\n",
       "      <th>probas</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.855460</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9415586552948044, 1.0, 0.6248227956728621]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.798582</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9294322124011538, 1.0, 0.46631374746266774]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>None</td>\n",
       "      <td>NaN</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.783383</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.8360490328877616, 1.0, 0.5140987736785823]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.984085</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9852943113674895, 1.0, 0.9669618390243637]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.834089</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9740126783506491, 1.0, 0.5282558161717601]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>496</th>\n",
       "      <td>None</td>\n",
       "      <td>NaN</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497</th>\n",
       "      <td>None</td>\n",
       "      <td>NaN</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>498</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.740054</td>\n",
       "      <td>[svm_ir_clf]</td>\n",
       "      <td>[0.7400541593058131]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>499</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.837794</td>\n",
       "      <td>[svm_ir_clf, best_logi]</td>\n",
       "      <td>[0.8109765254551804, 0.8646112118577693]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>500 rows × 4 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "        predict  mean_proba                                     models  \\\n",
       "0    IRRELEVANT    0.855460  [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "1    IRRELEVANT    0.798582  [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "2          None         NaN                                       None   \n",
       "3    IRRELEVANT    0.783383  [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "4    IRRELEVANT    0.984085  [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "..          ...         ...                                        ...   \n",
       "495  IRRELEVANT    0.834089  [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "496        None         NaN                                       None   \n",
       "497        None         NaN                                       None   \n",
       "498  IRRELEVANT    0.740054                               [svm_ir_clf]   \n",
       "499  IRRELEVANT    0.837794                    [svm_ir_clf, best_logi]   \n",
       "\n",
       "                                             probas  \n",
       "0     [0.9415586552948044, 1.0, 0.6248227956728621]  \n",
       "1    [0.9294322124011538, 1.0, 0.46631374746266774]  \n",
       "2                                              None  \n",
       "3     [0.8360490328877616, 1.0, 0.5140987736785823]  \n",
       "4     [0.9852943113674895, 1.0, 0.9669618390243637]  \n",
       "..                                              ...  \n",
       "495   [0.9740126783506491, 1.0, 0.5282558161717601]  \n",
       "496                                            None  \n",
       "497                                            None  \n",
       "498                            [0.7400541593058131]  \n",
       "499        [0.8109765254551804, 0.8646112118577693]  \n",
       "\n",
       "[500 rows x 4 columns]"
      ]
     },
     "execution_count": 147,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "irre_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 148,
   "metadata": {},
   "outputs": [
    {
     "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>label</th>\n",
       "      <th>predict</th>\n",
       "      <th>mean_proba</th>\n",
       "      <th>models</th>\n",
       "      <th>probas</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.855460</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9415586552948044, 1.0, 0.6248227956728621]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.798582</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9294322124011538, 1.0, 0.46631374746266774]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>FOREX MARKETS</td>\n",
       "      <td>MONEY MARKETS</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>[clf_NB_2]</td>\n",
       "      <td>[0.9999999999997726]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.783383</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.8360490328877616, 1.0, 0.5140987736785823]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.984085</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9852943113674895, 1.0, 0.9669618390243637]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.834089</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9740126783506491, 1.0, 0.5282558161717601]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>496</th>\n",
       "      <td>SPORTS</td>\n",
       "      <td>SPORTS</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>[tfidf_logi_pipe, clf_NB_1, clf_NB_2, svm_prob...</td>\n",
       "      <td>[1.0, 1.0, 1.0, 0.9999999995165704, 0.99999995...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497</th>\n",
       "      <td>MONEY MARKETS</td>\n",
       "      <td>MONEY MARKETS</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>[clf_NB_2]</td>\n",
       "      <td>[1.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>498</th>\n",
       "      <td>SHARE LISTINGS</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.740054</td>\n",
       "      <td>[svm_ir_clf]</td>\n",
       "      <td>[0.7400541593058131]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>499</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.837794</td>\n",
       "      <td>[svm_ir_clf, best_logi]</td>\n",
       "      <td>[0.8109765254551804, 0.8646112118577693]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>500 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "              label        predict  mean_proba  \\\n",
       "0        IRRELEVANT     IRRELEVANT    0.855460   \n",
       "1        IRRELEVANT     IRRELEVANT    0.798582   \n",
       "2     FOREX MARKETS  MONEY MARKETS    1.000000   \n",
       "3        IRRELEVANT     IRRELEVANT    0.783383   \n",
       "4        IRRELEVANT     IRRELEVANT    0.984085   \n",
       "..              ...            ...         ...   \n",
       "495      IRRELEVANT     IRRELEVANT    0.834089   \n",
       "496          SPORTS         SPORTS    1.000000   \n",
       "497   MONEY MARKETS  MONEY MARKETS    1.000000   \n",
       "498  SHARE LISTINGS     IRRELEVANT    0.740054   \n",
       "499      IRRELEVANT     IRRELEVANT    0.837794   \n",
       "\n",
       "                                                models  \\\n",
       "0            [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "1            [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "2                                           [clf_NB_2]   \n",
       "3            [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "4            [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "..                                                 ...   \n",
       "495          [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "496  [tfidf_logi_pipe, clf_NB_1, clf_NB_2, svm_prob...   \n",
       "497                                         [clf_NB_2]   \n",
       "498                                       [svm_ir_clf]   \n",
       "499                            [svm_ir_clf, best_logi]   \n",
       "\n",
       "                                                probas  \n",
       "0        [0.9415586552948044, 1.0, 0.6248227956728621]  \n",
       "1       [0.9294322124011538, 1.0, 0.46631374746266774]  \n",
       "2                                 [0.9999999999997726]  \n",
       "3        [0.8360490328877616, 1.0, 0.5140987736785823]  \n",
       "4        [0.9852943113674895, 1.0, 0.9669618390243637]  \n",
       "..                                                 ...  \n",
       "495      [0.9740126783506491, 1.0, 0.5282558161717601]  \n",
       "496  [1.0, 1.0, 1.0, 0.9999999995165704, 0.99999995...  \n",
       "497                                              [1.0]  \n",
       "498                               [0.7400541593058131]  \n",
       "499           [0.8109765254551804, 0.8646112118577693]  \n",
       "\n",
       "[500 rows x 5 columns]"
      ]
     },
     "execution_count": 148,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "res_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classification Report:\n",
      "                                  precision    recall  f1-score   support\n",
      "\n",
      "      ARTS CULTURE ENTERTAINMENT       0.33      0.67      0.44         3\n",
      "BIOGRAPHIES PERSONALITIES PEOPLE       0.62      0.33      0.43        15\n",
      "                         DEFENCE       1.00      0.23      0.38        13\n",
      "                DOMESTIC MARKETS       0.00      0.00      0.00         2\n",
      "                   FOREX MARKETS       0.61      0.77      0.68        48\n",
      "                          HEALTH       0.78      0.50      0.61        14\n",
      "                      IRRELEVANT       0.85      0.92      0.88       266\n",
      "                   MONEY MARKETS       0.67      0.59      0.63        69\n",
      "          SCIENCE AND TECHNOLOGY       0.00      0.00      0.00         3\n",
      "                  SHARE LISTINGS       0.00      0.00      0.00         7\n",
      "                          SPORTS       0.95      0.97      0.96        60\n",
      "\n",
      "                        accuracy                           0.80       500\n",
      "                       macro avg       0.53      0.45      0.46       500\n",
      "                    weighted avg       0.78      0.80      0.78       500\n",
      "\n",
      "[[  2   0   0   0   0   0   1   0   0   0   0]\n",
      " [  3   5   0   0   0   0   7   0   0   0   0]\n",
      " [  0   0   3   0   0   0  10   0   0   0   0]\n",
      " [  0   0   0   0   0   0   2   0   0   0   0]\n",
      " [  0   0   0   0  37   0   3   7   0   1   0]\n",
      " [  0   0   0   0   0   7   6   0   1   0   0]\n",
      " [  0   1   0   0   3   1 245  13   0   0   3]\n",
      " [  0   0   0   0  21   0   7  41   0   0   0]\n",
      " [  1   0   0   0   0   1   1   0   0   0   0]\n",
      " [  0   0   0   0   0   0   7   0   0   0   0]\n",
      " [  0   2   0   0   0   0   0   0   0   0  58]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 936x504 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "DEFENCE accuarcy is:\t\t\t\t 0.23076923076923078\n",
      "DOMESTIC MARKETS accuarcy is:\t\t\t\t 0.0\n",
      "MONEY MARKETS accuarcy is:\t\t\t\t 0.5942028985507246\n",
      "FOREX MARKETS accuarcy is:\t\t\t\t 0.7708333333333334\n",
      "SPORTS accuarcy is:\t\t\t\t 0.9666666666666667\n",
      "SHARE LISTINGS accuarcy is:\t\t\t\t 0.0\n",
      "BIOGRAPHIES PERSONALITIES PEOPLE accuarcy is:\t\t\t\t 0.3333333333333333\n",
      "HEALTH accuarcy is:\t\t\t\t 0.5\n",
      "ARTS CULTURE ENTERTAINMENT accuarcy is:\t\t\t\t 0.6666666666666666\n",
      "SCIENCE AND TECHNOLOGY accuarcy is:\t\t\t\t 0.0\n",
      "IRRELEVANT accuarcy is:\t\t\t\t 0.9210526315789473\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "report(res_df['label'], res_df['predict'])\n",
    "evaluate(res_df, 'label', 'predict')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 119,
   "metadata": {},
   "outputs": [],
   "source": [
    "# recommandation\n",
    "with open('recommandation_5.txt', 'w') as f:\n",
    "    f.write('3 level_1 add_weight all\\n')\n",
    "    for topic in topic_list:\n",
    "        if topic != \"irrelavant\":\n",
    "            df_filter = res_df[res_df['predict'] == topic]\n",
    "            f.write(topic)\n",
    "            f.write('\\n')\n",
    "            f.write(str(df_filter.sort_values(by=\"mean_proba\")['label'].head(10)))\n",
    "            f.write('\\n')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "metadata": {},
   "outputs": [],
   "source": [
    "# recommandation\n",
    "with open('recommandation_6.txt', 'w') as f:\n",
    "    f.write('1 level_1 add_weight all\\n')\n",
    "    for topic in topic_list:\n",
    "        if topic != \"irrelavant\":\n",
    "            df_filter = res_df[res_df['predict'] == topic]\n",
    "            f.write(topic)\n",
    "            f.write('\\n')\n",
    "            f.write(str(df_filter.sort_values(by=\"mean_proba\")['label'].head(10)))\n",
    "            f.write('\\n')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'ARTS CULTURE ENTERTAINMENT': 3,\n",
       " 'BIOGRAPHIES PERSONALITIES PEOPLE': 2,\n",
       " 'DEFENCE': 4,\n",
       " 'DOMESTIC MARKETS': 5,\n",
       " 'FOREX MARKETS': 4,\n",
       " 'HEALTH': 2,\n",
       " 'IRRELEVANT': 1,\n",
       " 'MONEY MARKETS': 1,\n",
       " 'SCIENCE AND TECHNOLOGY': 1,\n",
       " 'SHARE LISTINGS': 4,\n",
       " 'SPORTS': 8}"
      ]
     },
     "execution_count": 126,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "topic_count"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "metadata": {},
   "outputs": [],
   "source": [
    "# recommandation\n",
    "with open('recommandation_7.txt', 'w') as f:\n",
    "    f.write('3 level_1 times_weight all\\n')\n",
    "    for topic in topic_list:\n",
    "        if topic != \"irrelavant\":\n",
    "            df_filter = res_df[res_df['predict'] == topic]\n",
    "            f.write(topic)\n",
    "            f.write('\\n')\n",
    "            f.write(str(df_filter.sort_values(by=\"mean_proba\")['label'].head(10)))\n",
    "            f.write('\\n')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "metadata": {},
   "outputs": [],
   "source": [
    "# recommandation\n",
    "with open('recommandation_8.txt', 'w') as f:\n",
    "    f.write('3 level_1 times_weight all special level 1\\n')\n",
    "    for topic in topic_list:\n",
    "        if topic != \"irrelavant\":\n",
    "            df_filter = res_df[res_df['predict'] == topic]\n",
    "            f.write(topic)\n",
    "            f.write('\\n')\n",
    "            f.write(str(df_filter.sort_values(by=\"mean_proba\")['label'].head(10)))\n",
    "            f.write('\\n')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 151,
   "metadata": {},
   "outputs": [
    {
     "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>Unnamed: 0</th>\n",
       "      <th>article_number</th>\n",
       "      <th>article_words</th>\n",
       "      <th>topic</th>\n",
       "      <th>key_word_10</th>\n",
       "      <th>key_word_20</th>\n",
       "      <th>key_word_50</th>\n",
       "      <th>key_word_100</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>9501</td>\n",
       "      <td>world,complet,pharmaceut,tianjin,tianjin,chin,...</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>world,chin,chin,chin,chin,chin,chin,million,mi...</td>\n",
       "      <td>world,chin,chin,chin,chin,chin,chin,industr,gr...</td>\n",
       "      <td>world,chin,chin,chin,chin,chin,chin,industr,gr...</td>\n",
       "      <td>world,chin,chin,chin,chin,chin,chin,city,indus...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>9502</td>\n",
       "      <td>copy,sunday,weekend,ec,friday,eu,includ,limit,...</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>report</td>\n",
       "      <td>eu,includ,report</td>\n",
       "      <td>sunday,friday,eu,includ,european,report,saturday</td>\n",
       "      <td>sunday,friday,eu,includ,european,commun,servic...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>9503</td>\n",
       "      <td>heavy,heavy,gabriel,morn,morn,equit,cent,cent,...</td>\n",
       "      <td>FOREX MARKETS</td>\n",
       "      <td>stock,stock,stock,rate,rate,rate,day,day,day,p...</td>\n",
       "      <td>cent,cent,cent,stock,stock,stock,rate,rate,rat...</td>\n",
       "      <td>cent,cent,cent,stock,stock,stock,rate,rate,rat...</td>\n",
       "      <td>cent,cent,cent,recent,stock,stock,stock,rate,r...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>9504</td>\n",
       "      <td>research,jess,hit,anticip,comput,comput,comput...</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>day,work,offic,year,year,year,year,percent,per...</td>\n",
       "      <td>research,industr,industr,industr,industr,group...</td>\n",
       "      <td>research,comput,comput,comput,comput,comput,co...</td>\n",
       "      <td>research,hit,comput,comput,comput,comput,compu...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>9505</td>\n",
       "      <td>provid,provid,luxembourg,court,court,case,opin...</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>stat,stat,stat,stat,stat,stat,stat,bank,year,r...</td>\n",
       "      <td>case,end,unit,unit,unit,unit,stat,stat,stat,st...</td>\n",
       "      <td>court,court,case,end,unit,unit,unit,unit,stat,...</td>\n",
       "      <td>court,court,case,rule,rule,law,end,unit,unit,u...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495</th>\n",
       "      <td>495</td>\n",
       "      <td>9996</td>\n",
       "      <td>symantec,soon,interfac,provid,provid,comput,co...</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>year,year,year,compan,compan,system,produc,mar...</td>\n",
       "      <td>end,end,unit,unit,unit,includ,includ,year,year...</td>\n",
       "      <td>comput,comput,comput,comput,end,end,unit,unit,...</td>\n",
       "      <td>comput,comput,comput,comput,end,end,unit,unit,...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>496</th>\n",
       "      <td>496</td>\n",
       "      <td>9997</td>\n",
       "      <td>hit,hit,hit,hit,hit,hit,hit,hit,hit,beat,beat,...</td>\n",
       "      <td>SPORTS</td>\n",
       "      <td>scor,scor,scor,scor,scor,scor,scor,play,work,w...</td>\n",
       "      <td>beat,beat,scor,scor,scor,scor,scor,scor,scor,e...</td>\n",
       "      <td>beat,beat,scor,scor,scor,scor,scor,scor,scor,e...</td>\n",
       "      <td>hit,hit,hit,hit,hit,hit,hit,hit,hit,beat,beat,...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497</th>\n",
       "      <td>497</td>\n",
       "      <td>9998</td>\n",
       "      <td>cent,cent,cent,cent,match,declin,act,rate,rate...</td>\n",
       "      <td>MONEY MARKETS</td>\n",
       "      <td>match,rate,rate,rate,rate,rate,rate,day,day,ba...</td>\n",
       "      <td>cent,cent,cent,cent,match,act,rate,rate,rate,r...</td>\n",
       "      <td>cent,cent,cent,cent,match,act,rate,rate,rate,r...</td>\n",
       "      <td>cent,cent,cent,cent,match,act,rate,rate,rate,r...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>498</th>\n",
       "      <td>498</td>\n",
       "      <td>9999</td>\n",
       "      <td>cnmv,cnmv,stock,count,count,week,group,friday,...</td>\n",
       "      <td>SHARE LISTINGS</td>\n",
       "      <td>stock,bank,shar,percent,market,trad</td>\n",
       "      <td>stock,week,group,bank,shar,percent,suspend,sus...</td>\n",
       "      <td>stock,week,group,friday,franc,bank,shar,bid,bi...</td>\n",
       "      <td>stock,week,group,friday,franc,bank,shar,bid,bi...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>499</th>\n",
       "      <td>499</td>\n",
       "      <td>10000</td>\n",
       "      <td>matthey,matthey,metal,hit,morn,widen,widen,cen...</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>day,play,year,year,lead,report,deal,deal,deal,...</td>\n",
       "      <td>cent,day,play,year,year,lead,report,deal,deal,...</td>\n",
       "      <td>cent,day,play,profit,hold,made,early,year,year...</td>\n",
       "      <td>hit,cent,gmt,term,fall,day,fix,fix,fix,fix,pla...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>500 rows × 8 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     Unnamed: 0  article_number  \\\n",
       "0             0            9501   \n",
       "1             1            9502   \n",
       "2             2            9503   \n",
       "3             3            9504   \n",
       "4             4            9505   \n",
       "..          ...             ...   \n",
       "495         495            9996   \n",
       "496         496            9997   \n",
       "497         497            9998   \n",
       "498         498            9999   \n",
       "499         499           10000   \n",
       "\n",
       "                                         article_words           topic  \\\n",
       "0    world,complet,pharmaceut,tianjin,tianjin,chin,...      IRRELEVANT   \n",
       "1    copy,sunday,weekend,ec,friday,eu,includ,limit,...      IRRELEVANT   \n",
       "2    heavy,heavy,gabriel,morn,morn,equit,cent,cent,...   FOREX MARKETS   \n",
       "3    research,jess,hit,anticip,comput,comput,comput...      IRRELEVANT   \n",
       "4    provid,provid,luxembourg,court,court,case,opin...      IRRELEVANT   \n",
       "..                                                 ...             ...   \n",
       "495  symantec,soon,interfac,provid,provid,comput,co...      IRRELEVANT   \n",
       "496  hit,hit,hit,hit,hit,hit,hit,hit,hit,beat,beat,...          SPORTS   \n",
       "497  cent,cent,cent,cent,match,declin,act,rate,rate...   MONEY MARKETS   \n",
       "498  cnmv,cnmv,stock,count,count,week,group,friday,...  SHARE LISTINGS   \n",
       "499  matthey,matthey,metal,hit,morn,widen,widen,cen...      IRRELEVANT   \n",
       "\n",
       "                                           key_word_10  \\\n",
       "0    world,chin,chin,chin,chin,chin,chin,million,mi...   \n",
       "1                                               report   \n",
       "2    stock,stock,stock,rate,rate,rate,day,day,day,p...   \n",
       "3    day,work,offic,year,year,year,year,percent,per...   \n",
       "4    stat,stat,stat,stat,stat,stat,stat,bank,year,r...   \n",
       "..                                                 ...   \n",
       "495  year,year,year,compan,compan,system,produc,mar...   \n",
       "496  scor,scor,scor,scor,scor,scor,scor,play,work,w...   \n",
       "497  match,rate,rate,rate,rate,rate,rate,day,day,ba...   \n",
       "498                stock,bank,shar,percent,market,trad   \n",
       "499  day,play,year,year,lead,report,deal,deal,deal,...   \n",
       "\n",
       "                                           key_word_20  \\\n",
       "0    world,chin,chin,chin,chin,chin,chin,industr,gr...   \n",
       "1                                     eu,includ,report   \n",
       "2    cent,cent,cent,stock,stock,stock,rate,rate,rat...   \n",
       "3    research,industr,industr,industr,industr,group...   \n",
       "4    case,end,unit,unit,unit,unit,stat,stat,stat,st...   \n",
       "..                                                 ...   \n",
       "495  end,end,unit,unit,unit,includ,includ,year,year...   \n",
       "496  beat,beat,scor,scor,scor,scor,scor,scor,scor,e...   \n",
       "497  cent,cent,cent,cent,match,act,rate,rate,rate,r...   \n",
       "498  stock,week,group,bank,shar,percent,suspend,sus...   \n",
       "499  cent,day,play,year,year,lead,report,deal,deal,...   \n",
       "\n",
       "                                           key_word_50  \\\n",
       "0    world,chin,chin,chin,chin,chin,chin,industr,gr...   \n",
       "1     sunday,friday,eu,includ,european,report,saturday   \n",
       "2    cent,cent,cent,stock,stock,stock,rate,rate,rat...   \n",
       "3    research,comput,comput,comput,comput,comput,co...   \n",
       "4    court,court,case,end,unit,unit,unit,unit,stat,...   \n",
       "..                                                 ...   \n",
       "495  comput,comput,comput,comput,end,end,unit,unit,...   \n",
       "496  beat,beat,scor,scor,scor,scor,scor,scor,scor,e...   \n",
       "497  cent,cent,cent,cent,match,act,rate,rate,rate,r...   \n",
       "498  stock,week,group,friday,franc,bank,shar,bid,bi...   \n",
       "499  cent,day,play,profit,hold,made,early,year,year...   \n",
       "\n",
       "                                          key_word_100  \n",
       "0    world,chin,chin,chin,chin,chin,chin,city,indus...  \n",
       "1    sunday,friday,eu,includ,european,commun,servic...  \n",
       "2    cent,cent,cent,recent,stock,stock,stock,rate,r...  \n",
       "3    research,hit,comput,comput,comput,comput,compu...  \n",
       "4    court,court,case,rule,rule,law,end,unit,unit,u...  \n",
       "..                                                 ...  \n",
       "495  comput,comput,comput,comput,end,end,unit,unit,...  \n",
       "496  hit,hit,hit,hit,hit,hit,hit,hit,hit,beat,beat,...  \n",
       "497  cent,cent,cent,cent,match,act,rate,rate,rate,r...  \n",
       "498  stock,week,group,friday,franc,bank,shar,bid,bi...  \n",
       "499  hit,cent,gmt,term,fall,day,fix,fix,fix,fix,pla...  \n",
       "\n",
       "[500 rows x 8 columns]"
      ]
     },
     "execution_count": 151,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 152,
   "metadata": {},
   "outputs": [
    {
     "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>label</th>\n",
       "      <th>predict</th>\n",
       "      <th>mean_proba</th>\n",
       "      <th>models</th>\n",
       "      <th>probas</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>article_number</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>9501</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.855460</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9415586552948044, 1.0, 0.6248227956728621]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9502</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.798582</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9294322124011538, 1.0, 0.46631374746266774]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9503</th>\n",
       "      <td>FOREX MARKETS</td>\n",
       "      <td>MONEY MARKETS</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>[clf_NB_2]</td>\n",
       "      <td>[0.9999999999997726]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9504</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.783383</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.8360490328877616, 1.0, 0.5140987736785823]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9505</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.984085</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9852943113674895, 1.0, 0.9669618390243637]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9996</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.834089</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9740126783506491, 1.0, 0.5282558161717601]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9997</th>\n",
       "      <td>SPORTS</td>\n",
       "      <td>SPORTS</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>[tfidf_logi_pipe, clf_NB_1, clf_NB_2, svm_prob...</td>\n",
       "      <td>[1.0, 1.0, 1.0, 0.9999999995165704, 0.99999995...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9998</th>\n",
       "      <td>MONEY MARKETS</td>\n",
       "      <td>MONEY MARKETS</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>[clf_NB_2]</td>\n",
       "      <td>[1.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9999</th>\n",
       "      <td>SHARE LISTINGS</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.740054</td>\n",
       "      <td>[svm_ir_clf]</td>\n",
       "      <td>[0.7400541593058131]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10000</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.837794</td>\n",
       "      <td>[svm_ir_clf, best_logi]</td>\n",
       "      <td>[0.8109765254551804, 0.8646112118577693]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>500 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                         label        predict  mean_proba  \\\n",
       "article_number                                              \n",
       "9501                IRRELEVANT     IRRELEVANT    0.855460   \n",
       "9502                IRRELEVANT     IRRELEVANT    0.798582   \n",
       "9503             FOREX MARKETS  MONEY MARKETS    1.000000   \n",
       "9504                IRRELEVANT     IRRELEVANT    0.783383   \n",
       "9505                IRRELEVANT     IRRELEVANT    0.984085   \n",
       "...                        ...            ...         ...   \n",
       "9996                IRRELEVANT     IRRELEVANT    0.834089   \n",
       "9997                    SPORTS         SPORTS    1.000000   \n",
       "9998             MONEY MARKETS  MONEY MARKETS    1.000000   \n",
       "9999            SHARE LISTINGS     IRRELEVANT    0.740054   \n",
       "10000               IRRELEVANT     IRRELEVANT    0.837794   \n",
       "\n",
       "                                                           models  \\\n",
       "article_number                                                      \n",
       "9501                    [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "9502                    [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "9503                                                   [clf_NB_2]   \n",
       "9504                    [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "9505                    [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "...                                                           ...   \n",
       "9996                    [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "9997            [tfidf_logi_pipe, clf_NB_1, clf_NB_2, svm_prob...   \n",
       "9998                                                   [clf_NB_2]   \n",
       "9999                                                 [svm_ir_clf]   \n",
       "10000                                     [svm_ir_clf, best_logi]   \n",
       "\n",
       "                                                           probas  \n",
       "article_number                                                     \n",
       "9501                [0.9415586552948044, 1.0, 0.6248227956728621]  \n",
       "9502               [0.9294322124011538, 1.0, 0.46631374746266774]  \n",
       "9503                                         [0.9999999999997726]  \n",
       "9504                [0.8360490328877616, 1.0, 0.5140987736785823]  \n",
       "9505                [0.9852943113674895, 1.0, 0.9669618390243637]  \n",
       "...                                                           ...  \n",
       "9996                [0.9740126783506491, 1.0, 0.5282558161717601]  \n",
       "9997            [1.0, 1.0, 1.0, 0.9999999995165704, 0.99999995...  \n",
       "9998                                                        [1.0]  \n",
       "9999                                         [0.7400541593058131]  \n",
       "10000                    [0.8109765254551804, 0.8646112118577693]  \n",
       "\n",
       "[500 rows x 5 columns]"
      ]
     },
     "execution_count": 152,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "res_df.set_index(df.article_number)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 153,
   "metadata": {},
   "outputs": [
    {
     "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>label</th>\n",
       "      <th>predict</th>\n",
       "      <th>mean_proba</th>\n",
       "      <th>models</th>\n",
       "      <th>probas</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.855460</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9415586552948044, 1.0, 0.6248227956728621]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.798582</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9294322124011538, 1.0, 0.46631374746266774]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>FOREX MARKETS</td>\n",
       "      <td>MONEY MARKETS</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>[clf_NB_2]</td>\n",
       "      <td>[0.9999999999997726]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.783383</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.8360490328877616, 1.0, 0.5140987736785823]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.984085</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9852943113674895, 1.0, 0.9669618390243637]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.834089</td>\n",
       "      <td>[svm_ir_clf, rel_clf_tfidf_NB, best_logi]</td>\n",
       "      <td>[0.9740126783506491, 1.0, 0.5282558161717601]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>496</th>\n",
       "      <td>SPORTS</td>\n",
       "      <td>SPORTS</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>[tfidf_logi_pipe, clf_NB_1, clf_NB_2, svm_prob...</td>\n",
       "      <td>[1.0, 1.0, 1.0, 0.9999999995165704, 0.99999995...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>497</th>\n",
       "      <td>MONEY MARKETS</td>\n",
       "      <td>MONEY MARKETS</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>[clf_NB_2]</td>\n",
       "      <td>[1.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>498</th>\n",
       "      <td>SHARE LISTINGS</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.740054</td>\n",
       "      <td>[svm_ir_clf]</td>\n",
       "      <td>[0.7400541593058131]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>499</th>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>IRRELEVANT</td>\n",
       "      <td>0.837794</td>\n",
       "      <td>[svm_ir_clf, best_logi]</td>\n",
       "      <td>[0.8109765254551804, 0.8646112118577693]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>500 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "              label        predict  mean_proba  \\\n",
       "0        IRRELEVANT     IRRELEVANT    0.855460   \n",
       "1        IRRELEVANT     IRRELEVANT    0.798582   \n",
       "2     FOREX MARKETS  MONEY MARKETS    1.000000   \n",
       "3        IRRELEVANT     IRRELEVANT    0.783383   \n",
       "4        IRRELEVANT     IRRELEVANT    0.984085   \n",
       "..              ...            ...         ...   \n",
       "495      IRRELEVANT     IRRELEVANT    0.834089   \n",
       "496          SPORTS         SPORTS    1.000000   \n",
       "497   MONEY MARKETS  MONEY MARKETS    1.000000   \n",
       "498  SHARE LISTINGS     IRRELEVANT    0.740054   \n",
       "499      IRRELEVANT     IRRELEVANT    0.837794   \n",
       "\n",
       "                                                models  \\\n",
       "0            [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "1            [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "2                                           [clf_NB_2]   \n",
       "3            [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "4            [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "..                                                 ...   \n",
       "495          [svm_ir_clf, rel_clf_tfidf_NB, best_logi]   \n",
       "496  [tfidf_logi_pipe, clf_NB_1, clf_NB_2, svm_prob...   \n",
       "497                                         [clf_NB_2]   \n",
       "498                                       [svm_ir_clf]   \n",
       "499                            [svm_ir_clf, best_logi]   \n",
       "\n",
       "                                                probas  \n",
       "0        [0.9415586552948044, 1.0, 0.6248227956728621]  \n",
       "1       [0.9294322124011538, 1.0, 0.46631374746266774]  \n",
       "2                                 [0.9999999999997726]  \n",
       "3        [0.8360490328877616, 1.0, 0.5140987736785823]  \n",
       "4        [0.9852943113674895, 1.0, 0.9669618390243637]  \n",
       "..                                                 ...  \n",
       "495      [0.9740126783506491, 1.0, 0.5282558161717601]  \n",
       "496  [1.0, 1.0, 1.0, 0.9999999995165704, 0.99999995...  \n",
       "497                                              [1.0]  \n",
       "498                               [0.7400541593058131]  \n",
       "499           [0.8109765254551804, 0.8646112118577693]  \n",
       "\n",
       "[500 rows x 5 columns]"
      ]
     },
     "execution_count": 153,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "res_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.7.4 64-bit",
   "language": "python",
   "name": "python37464bit54f3a551cc7b4ff589218c6160942c64"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}