Task.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Домашнее задание по обработке текстов\n",
    "\n",
    "## Предсказание цены акции по экономическим новостям\n",
    "\n",
    "Входные данные:\n",
    "* Новости о компании \"Газпром\", начиная с 2010 года\n",
    "* Стоимость акций компании \"Газпром\" на ММВБ, начиная с 2010 года\n",
    "    * цена открытия (Open)\n",
    "    * цена закрытия (ClosingPrice)\n",
    "    * максимальная цена за день (DailyHigh)\n",
    "    * минимальная цена за день (DailyLow) \n",
    "    * объем бумаг (VolumePcs)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Populating the interactive namespace from numpy and matplotlib\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import seaborn as sns\n",
    "import re\n",
    "%pylab inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "#импортируем данные с переводом текста в дату и разделения десятичной части запятой.\n",
    "#в качестве индекса удобно использовать дату\n",
    "df = pd.read_csv('texts.csv', parse_dates=[0], index_col=0, converters={1:lambda x: x.replace('\\xa0', ' ')})\n",
    "pr_all = pd.read_csv('gazprom_prices.csv', sep=';', parse_dates=[0], decimal=',', index_col=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Оставляем только даты, когда были и новости и данные по стоимости акций\n",
    "df = df[df.index.isin(pr_all.index)]\n",
    "pr_all = pr_all[pr_all.index.isin(df.index)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1159"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Итого получаем дней для анализа:\n",
    "len(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(                                                         text\n",
       " date                                                         \n",
       " 2017-09-11  Компания рассчитывает на решение по газовому с...\n",
       " 2017-08-11   Как и предполагал “Ъ”, «Газпром», воспользова...\n",
       " 2017-01-11   Новая редакция американских санкций ставит по...\n",
       " 2017-10-30   Как стало известно “Ъ”, известный на рынке ри...\n",
       " 2017-10-23   НОВАТЭК, который через пять лет собирается за...,\n",
       "               Open  ClosingPrice  DailyHigh  DailyLow  VolumePcs\n",
       " Date                                                            \n",
       " 2017-09-11  132.31        131.50     132.82    131.14   33869650\n",
       " 2017-08-11  132.00        132.30     133.94    131.58   39381960\n",
       " 2017-01-11  126.40        126.50     126.89    125.97   18232550\n",
       " 2017-10-30  125.96        125.98     126.93    125.53   19263340\n",
       " 2017-10-23  127.05        126.80     127.47    126.37   17308800)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head(), pr_all.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Часть 1. Вводная\n",
    "\n",
    "Проведите предобработку текстов: если считаете нужным, выполните токенизацию, приведение к нижнему регистру, лемматизацию и/или стемминг. Ответьте на следующие вопросы:\n",
    "* Есть ли корреляция между средней длинной текста за день и ценой закрытия?\n",
    "* Есть ли корреляция между количеством упоминаний Алексея Миллера  и ценой закрытия? Учтите разные варианты написания имени.\n",
    "* Упоминаний какого газопровода в статьях больше: \n",
    "    * \"северный поток\"\n",
    "    * \"турецкий поток\"?\n",
    "* О каких санкциях пишут в статьях?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Есть ли корреляция между средней длинной текста за день и ценой закрытия?##### "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#Посчитаем длинну новостей\n",
    "df['text_len'] = df['text'].apply(len)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.014409399726613269"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.corrcoef(pr_all.ClosingPrice, df.text_len)[0,1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Коэффициент корреляции Пирсона меньше 0.3 \n",
    "**Вывод:** корреляция слабая или отсутствует"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<seaborn.axisgrid.FacetGrid at 0x1f3bd3daf98>"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWAAAAFgCAYAAACFYaNMAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1f3bd43a048>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Построим график для наглядности\n",
    "corr = pd.DataFrame()\n",
    "corr['Text_len']=df.text_len\n",
    "corr['Price']=pr_all.ClosingPrice\n",
    "sns.lmplot('Text_len',\n",
    "           'Price', \n",
    "           data=corr, \n",
    "           fit_reg=False, \n",
    "           scatter_kws={\"marker\": \"D\", \n",
    "                        \"s\": 10}) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Есть ли корреляция между количеством упоминаний Алексея Миллера и ценой закрытия? #### "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Посчитаем количество упоминаний Алексея Миллера в новостях.\n",
    "df['Miller']=df['text'].apply(lambda text: len(re.findall('[Мм]иллер', text)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.010635076467578011"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.corrcoef(pr_all.ClosingPrice, df.Miller)[0,1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Коэффициент корреляции Пирсона меньше 0.3 **Вывод**: корреляция слабая или отсутствует"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<seaborn.axisgrid.FacetGrid at 0x1f3bd3da278>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWAAAAFgCAYAAACFYaNMAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1f3bd3da1d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Построим график для наглядности\n",
    "m_corr = pd.DataFrame()\n",
    "m_corr['Miller_counts']=df.Miller\n",
    "m_corr['Price']=pr_all.ClosingPrice\n",
    "sns.lmplot('Miller_counts',\n",
    "           'Price', \n",
    "           data=m_corr, \n",
    "           fit_reg=False, \n",
    "           scatter_kws={\"marker\": \"D\", \n",
    "                        \"s\": 10}) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Упоминаний какого газопровода в статьях больше?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Упоминаний Турецкого потока: 39\n",
      "Упоминаний Северного потока: 15\n"
     ]
    }
   ],
   "source": [
    "#Загрузим данные еще раз, чтобы иметь новости за все даты\n",
    "df1 = pd.read_csv('texts.csv', parse_dates=[0], index_col=0, converters={1:lambda x: x.replace('\\xa0', ' ')})\n",
    "\n",
    "north = sum(df1['text'].apply(lambda text: len(re.findall('[Сс]еверн[а-я]{2,4} [Пп]оток[а-я]{0,3}', text))))\n",
    "turkish = sum(df1['text'].apply(lambda text: len(re.findall('[Тт]урецк[а-я]{2,4} [Пп]оток[а-я]{0,3}', text))))\n",
    "print(\"Упоминаний Турецкого потока: {}\\nУпоминаний Северного потока: {}\".format(turkish,north))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### О каких санкциях пишут в статьях?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#Лемматизация\n",
    "from pymorphy2 import MorphAnalyzer\n",
    "\n",
    "def lem(text):\n",
    "    m =  MorphAnalyzer()\n",
    "    return  [m.parse(word)[0].normal_form for word in text.split()]\n",
    "\n",
    "df1['lemmas']=df1['text'].apply(lem)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Оставим только новости о санкциях\n",
    "sanctions = df1['lemmas'][df1['lemmas'].apply(lambda x: 'санкция' in x)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Функция для поиска прилагательных перед словом \"санкция\"\n",
    "def sanc_a(text):\n",
    "    m= MorphAnalyzer()\n",
    "    word_index = [i-1 for i, x in enumerate(text) if x == \"санкция\"]\n",
    "    adj = []\n",
    "    for i in word_index:\n",
    "        if m.parse(text[i])[0].tag.POS in ['ADJF', 'ADJS']:\n",
    "            adj.append(text[i])\n",
    "    return adj"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'американский',\n",
       " 'западный',\n",
       " 'персональный',\n",
       " 'секторальный',\n",
       " 'штрафной',\n",
       " 'экономический'}"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "set(sum(sanctions.apply(sanc_a)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "В списке выше можно видеть о каких санкциях идет речь в новостях."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Часть 2. Классификационная\n",
    "Вам предстоит решить следующую задачу: по текстам новостей за день определить, вырастет или понизится цена закрытия.\n",
    "Для этого:\n",
    "* бинаризуйте признак \"цена закрытия\":  новый признак ClosingPrice_bin равен 1, если по сравнению со вчера цена не упала, и 0 – в обратном случаея;\n",
    "* составьте бучающее и тестовое множество: данные до начала 2016 года используются для обучения, данные с 2016 года и позже – для тестирования.\n",
    "\n",
    "Таким образом, в каждлый момент времени мы знаем: \n",
    "* ClosingPrice_bin – бинарый целевой признак\n",
    "* слова из статей, опубликованных в этот день – объясняющие признаки\n",
    "\n",
    "В этой части задания вам нужно сделать baseline алгоритм и попытаться его улучшить в следующей части. \n",
    "\n",
    "Используйте любой известный вам алгоритм классификации текстов для того, Используйте $tf-idf$ преобразование, сингулярное разложение, нормировку признакого пространства и любые другие техники обработки данных, которые вы считаете нужным. Используйте accuracy и F-measure для оценки качества классификации. Покажите, как  $tf-idf$ преобразование или сингулярное разложение или любая другая использованная вами техника влияет на качество классификации.\n",
    "Если у выбранного вами алгоритма есть гиперпараметры (например, $\\alpha$ в преобразовании Лапласа для метода наивного Байеса), покажите, как изменение гиперпараметра влияет на качество классификации."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Бинаризуйте признак \"цена закрытия\": новый признак ClosingPrice_bin равен 1, если по сравнению со вчера цена не упала, и 0 – в обратном случае#### "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Open</th>\n",
       "      <th>ClosingPrice</th>\n",
       "      <th>DailyHigh</th>\n",
       "      <th>DailyLow</th>\n",
       "      <th>VolumePcs</th>\n",
       "      <th>ClosingPrice_bin</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Date</th>\n",
       "      <th></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>2017-09-11</th>\n",
       "      <td>132.31</td>\n",
       "      <td>131.50</td>\n",
       "      <td>132.82</td>\n",
       "      <td>131.14</td>\n",
       "      <td>33869650</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2017-08-11</th>\n",
       "      <td>132.00</td>\n",
       "      <td>132.30</td>\n",
       "      <td>133.94</td>\n",
       "      <td>131.58</td>\n",
       "      <td>39381960</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2017-01-11</th>\n",
       "      <td>126.40</td>\n",
       "      <td>126.50</td>\n",
       "      <td>126.89</td>\n",
       "      <td>125.97</td>\n",
       "      <td>18232550</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2017-10-30</th>\n",
       "      <td>125.96</td>\n",
       "      <td>125.98</td>\n",
       "      <td>126.93</td>\n",
       "      <td>125.53</td>\n",
       "      <td>19263340</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2017-10-23</th>\n",
       "      <td>127.05</td>\n",
       "      <td>126.80</td>\n",
       "      <td>127.47</td>\n",
       "      <td>126.37</td>\n",
       "      <td>17308800</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              Open  ClosingPrice  DailyHigh  DailyLow  VolumePcs  \\\n",
       "Date                                                               \n",
       "2017-09-11  132.31        131.50     132.82    131.14   33869650   \n",
       "2017-08-11  132.00        132.30     133.94    131.58   39381960   \n",
       "2017-01-11  126.40        126.50     126.89    125.97   18232550   \n",
       "2017-10-30  125.96        125.98     126.93    125.53   19263340   \n",
       "2017-10-23  127.05        126.80     127.47    126.37   17308800   \n",
       "\n",
       "            ClosingPrice_bin  \n",
       "Date                          \n",
       "2017-09-11                 0  \n",
       "2017-08-11                 1  \n",
       "2017-01-11                 0  \n",
       "2017-10-30                 0  \n",
       "2017-10-23                 1  "
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pr_all['ClosingPrice_bin'] = pr_all.ClosingPrice >= pr_all.ClosingPrice.shift()\n",
    "pr_all['ClosingPrice_bin'] = pr_all['ClosingPrice_bin'].apply(lambda x: 1 if x == True else 0)\n",
    "pr_all.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Добавим леммы в обрезанную выборку\n",
    "df['lemmas']=df1['lemmas'][df1.index.isin(df.index)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Функция разбивки данных на обучающую и тестовую выборки\n",
    "def train_test_split_custom(X, y, year):\n",
    "    X_train = X[X.index.year < year]\n",
    "    X_test = X[X.index.year >= year]\n",
    "    y_train = y[y.index.isin(X_train.index)]\n",
    "    y_test = y[y.index.isin(X_test.index)]\n",
    "    return X_train, X_test, y_train, y_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = df['lemmas'].apply(\" \".join)\n",
    "y = pr_all['ClosingPrice_bin']\n",
    "X_train, X_test, y_train, y_test = train_test_split_custom(X, y, 2016)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.naive_bayes import MultinomialNB\n",
    "from sklearn.linear_model import SGDClassifier, LogisticRegression\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.decomposition import TruncatedSVD\n",
    "from sklearn.feature_extraction.text import *\n",
    "from sklearn.metrics import *\n",
    "from sklearn.pipeline import Pipeline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "stop_words = stopwords.words('russian')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Используем RandomForest без tf-idf"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Остальные параметры заранее подобраны для наибольшей точности"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 241,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Pipeline(memory=None,\n",
       "     steps=[('vect', CountVectorizer(analyzer='word', binary=False, decode_error='strict',\n",
       "        dtype=<class 'numpy.int64'>, encoding='utf-8', input='content',\n",
       "        lowercase=True, max_df=1.0, max_features=None, min_df=1,\n",
       "        ngram_range=(1, 2), preprocessor=None,\n",
       "        stop_words=['и', 'в', ...stimators=100, n_jobs=1,\n",
       "            oob_score=False, random_state=0, verbose=0, warm_start=False))])"
      ]
     },
     "execution_count": 241,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf = Pipeline([\n",
    "    ('vect', CountVectorizer(ngram_range=(1,2), stop_words=stop_words)),\n",
    "    #('tfidf', TfidfTransformer()),\n",
    "    ('reducer', TruncatedSVD(n_components=30)),\n",
    "    ('clf', RandomForestClassifier(random_state=0, n_estimators=100)),\n",
    "])\n",
    "\n",
    "\n",
    "clf.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 242,
   "metadata": {},
   "outputs": [],
   "source": [
    "predictions = clf.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 243,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Precision:   0.50\n",
      "Recall:   0.50\n",
      "F1-measure:   0.50\n",
      "Accuracy:   0.50\n",
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       0.44      0.49      0.46       112\n",
      "          1       0.57      0.51      0.54       146\n",
      "\n",
      "avg / total       0.51      0.50      0.51       258\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXEAAAEICAYAAACpqsStAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nl6TCjLgkFWbEJakwIy5JhRlxSSrs/wGMgddy7BElrgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1f40dfa46a0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(\"Precision: {0:6.2f}\".format(precision_score(y_test, predictions, average='macro')))\n",
    "print(\"Recall: {0:6.2f}\".format(recall_score(y_test, predictions, average='macro')))\n",
    "print(\"F1-measure: {0:6.2f}\".format(f1_score(y_test, predictions, average='macro')))\n",
    "print(\"Accuracy: {0:6.2f}\".format(accuracy_score(y_test, predictions)))\n",
    "print(classification_report(y_test, predictions))\n",
    "labels = clf.classes_\n",
    "sns.heatmap(data=confusion_matrix(y_test, predictions), annot=True, fmt=\"d\", cbar=False, xticklabels=labels, yticklabels=labels)\n",
    "plt.title(\"Confusion matrix\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Теперь его же но с tf-idf "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 238,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Pipeline(memory=None,\n",
       "     steps=[('vect', CountVectorizer(analyzer='word', binary=False, decode_error='strict',\n",
       "        dtype=<class 'numpy.int64'>, encoding='utf-8', input='content',\n",
       "        lowercase=True, max_df=1.0, max_features=None, min_df=1,\n",
       "        ngram_range=(1, 2), preprocessor=None,\n",
       "        stop_words=['и', 'в', ...stimators=100, n_jobs=1,\n",
       "            oob_score=False, random_state=0, verbose=0, warm_start=False))])"
      ]
     },
     "execution_count": 238,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf = Pipeline([\n",
    "    ('vect', CountVectorizer(ngram_range=(1,2), stop_words=stop_words)),\n",
    "    ('tfidf', TfidfTransformer()),\n",
    "    ('reducer', TruncatedSVD(n_components=30)),\n",
    "    ('clf', RandomForestClassifier(random_state=0, n_estimators=100)),\n",
    "])\n",
    "\n",
    "\n",
    "clf.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 239,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "predictions = clf.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 240,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Precision:   0.56\n",
      "Recall:   0.56\n",
      "F1-measure:   0.56\n",
      "Accuracy:   0.57\n",
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       0.51      0.46      0.49       112\n",
      "          1       0.62      0.66      0.64       146\n",
      "\n",
      "avg / total       0.57      0.57      0.57       258\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXEAAAEICAYAAACpqsStAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nGXFJSsyIS1JiRlySEvs/DD5apfoXCZwAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1f40cca7cf8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(\"Precision: {0:6.2f}\".format(precision_score(y_test, predictions, average='macro')))\n",
    "print(\"Recall: {0:6.2f}\".format(recall_score(y_test, predictions, average='macro')))\n",
    "print(\"F1-measure: {0:6.2f}\".format(f1_score(y_test, predictions, average='macro')))\n",
    "print(\"Accuracy: {0:6.2f}\".format(accuracy_score(y_test, predictions)))\n",
    "print(classification_report(y_test, predictions))\n",
    "labels = clf.classes_\n",
    "sns.heatmap(data=confusion_matrix(y_test, predictions), annot=True, fmt=\"d\", cbar=False, xticklabels=labels, yticklabels=labels)\n",
    "plt.title(\"Confusion matrix\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Выводы"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Несмотря на то, что добавление tf-idf преобразования несколько улучшает общую точность предсказаний, сами результаты оставляют желать лучшего. Реузультаты MultinomialNB и LogisticRegression оказались хуже, чем у RandomForest.    "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Часть 3. Творческая\n",
    "Придумайте и попытайтесь сделать еще что-нибудь, чтобы улучшить качество классификации. \n",
    "Направления развития:\n",
    "* Морфологический признаки: \n",
    "    * использовать в качестве признаков только существительные или только именованные сущности;\n",
    "* Модели скрытых тем:\n",
    "    * использовать в качестве признаков скрытые темы;\n",
    "    * использовать в качестве признаков динамические скрытые темы \n",
    "    пример тут: (https://github.com/RaRe-Technologies/gensim/blob/develop/docs/notebooks/dtm_example.ipynb)\n",
    "* Синтаксические признаки:\n",
    "    * использовать SOV-тройки в качестве признаков\n",
    "    * кластеризовать SOV-тройки по эмбеддингам глаголов (обученные word2vec модели можно скачать отсюда: (http://rusvectores.org/ru/models/) и использовать только центроиды кластеров в качестве признаков\n",
    "* что-нибудь еще     "
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}