hw4_part2_habr_popularity_ridge.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<center>\n",
    "<img src=\"../../img/ods_stickers.jpg\">\n",
    "## Открытый курс по машинному обучению\n",
    "</center>\n",
    "<center>Автор материала: Ольга Дайховская"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# <center> Домашнее задание № 4. Часть 2\n",
    "## <center>  Прогнозирование популярности статей на TechMedia с помощью линейных моделей\n",
    "\n",
    "**В задании Вам предлагается разобраться с тем, как работает TfidfVectorizer и DictVectorizer, затем обучить и настроить модель линейной регрессии Ridge на данных о публикациях на Хабрахабре. Пройдя все шаги, вы сможете получить бейзлайн для [соревнования](https://inclass.kaggle.com/c/howpop-habrahabr-favs-lognorm). \n",
    "Ответьте на все вопросы в этой тетрадке и заполните ответы в [гугл-форме](https://docs.google.com/forms/d/e/1FAIpQLSd8E7naU38ikc9-rd31KeVfez3emCo4Ok1WrpQTr-XCDelRiw/viewform).**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Описание соревнования**\n",
    "\n",
    "Предскажите, как много звездочек наберет статья, зная только ее текст и время публикации\n",
    "\n",
    "Необходимо предсказать популярность поста на Хабре по содержанию и времени публикации. Как известно, пользователи Хабра могут добавлять статьи к себе в избранное. Общее количество пользователей, которое это сделали отображается у статьи количеством звездочек. Будем считать, что число звездочек, поставленных статье, наиболее хорошо отражает ее популярность.\n",
    "\n",
    "Более формально, в качестве метрики популярности статьи будем использовать долю статей за последний месяц, у которых количество звездочек меньше чем у текущей статьи. А точнее, доле числа звездочек можно поставить в соответствие квантили стандартного распределения, таким образом получаем числовую характеристику популярности статьи. Популярность статьи 0 означает, что статья получила ровно столько звездочек, сколько в среднем получают статьи. И соответственно чем больше звездочек получила статья по сравнению со средним, тем выше это число."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Приступим:** импортируем необходимые библиотеки и скачаем данные"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn.feature_extraction import DictVectorizer\n",
    "\n",
    "#В зависимости от версии sklearn\n",
    "from sklearn.model_selection import train_test_split\n",
    "#from sklearn.cross_validation import train_test_split\n",
    "\n",
    "from sklearn.linear_model import Ridge\n",
    "from sklearn.metrics import mean_squared_error\n",
    "\n",
    "import numpy as np\n",
    "import scipy\n",
    "\n",
    "%matplotlib inline\n",
    "from matplotlib import pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "train_df = pd.read_csv('../../data/howpop_train.csv')\n",
    "test_df  = pd.read_csv('../../data/howpop_test.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>url</th>\n",
       "      <td>https://habrahabr.ru/post/18284/</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>domain</th>\n",
       "      <td>habrahabr.ru</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>post_id</th>\n",
       "      <td>18284</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>published</th>\n",
       "      <td>2008-01-01 18:19:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>author</th>\n",
       "      <td>@Tapac</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>flow</th>\n",
       "      <td>develop</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>polling</th>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>content_len</th>\n",
       "      <td>4305</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>title</th>\n",
       "      <td>Новогодний подарок блоггерам — WordPress 2.3.2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>comments</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>favs</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>views</th>\n",
       "      <td>236</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>votes_plus</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>votes_minus</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>views_lognorm</th>\n",
       "      <td>-0.792687</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>favs_lognorm</th>\n",
       "      <td>-1.34407</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>comments_lognorm</th>\n",
       "      <td>-2.43687</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                               0\n",
       "url                             https://habrahabr.ru/post/18284/\n",
       "domain                                              habrahabr.ru\n",
       "post_id                                                    18284\n",
       "published                                    2008-01-01 18:19:00\n",
       "author                                                    @Tapac\n",
       "flow                                                     develop\n",
       "polling                                                    False\n",
       "content_len                                                 4305\n",
       "title             Новогодний подарок блоггерам — WordPress 2.3.2\n",
       "comments                                                       0\n",
       "favs                                                           0\n",
       "views                                                        236\n",
       "votes_plus                                                     0\n",
       "votes_minus                                                    0\n",
       "views_lognorm                                          -0.792687\n",
       "favs_lognorm                                            -1.34407\n",
       "comments_lognorm                                        -2.43687"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df.head(1).T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((134137, 17), (3990, 9))"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df.shape, test_df.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Убедимся, что данные отсортированы по признаку `published`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAg0AAAFwCAYAAAA7TEgWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fd63649c9e8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "train_df['published'].apply(lambda ts: pd.to_datetime(ts).value).plot();"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Чтобы ответить на вопросы 1 и 2, можно использовать [pandas.DataFrame.corr()](http://pandas.pydata.org/pandas-docs/stable/generated/pandas.DataFrame.corr.html), [pandas.to_datetime()](http://pandas.pydata.org/pandas-docs/stable/generated/pandas.to_datetime.html) и [pandas.Series.value_counts()](http://pandas.pydata.org/pandas-docs/stable/generated/pandas.Series.value_counts.html)**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<font color='red'>Вопрос 1.</font> Есть ли в train_df признаки, корреляция между которыми больше 0.9? Обратите внимание, именно различные признаки - корреляция признака с самим собой естественно больше 0.9 :)\n",
    "- да\n",
    "- нет\n",
    "- не знаю"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>post_id</th>\n",
       "      <th>polling</th>\n",
       "      <th>content_len</th>\n",
       "      <th>comments</th>\n",
       "      <th>favs</th>\n",
       "      <th>views</th>\n",
       "      <th>votes_plus</th>\n",
       "      <th>votes_minus</th>\n",
       "      <th>views_lognorm</th>\n",
       "      <th>favs_lognorm</th>\n",
       "      <th>comments_lognorm</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>post_id</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.081628</td>\n",
       "      <td>0.241384</td>\n",
       "      <td>-0.147763</td>\n",
       "      <td>0.089313</td>\n",
       "      <td>0.132385</td>\n",
       "      <td>-0.187885</td>\n",
       "      <td>-0.235594</td>\n",
       "      <td>-0.010376</td>\n",
       "      <td>-0.005845</td>\n",
       "      <td>-0.001605</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>polling</th>\n",
       "      <td>0.081628</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>-0.003337</td>\n",
       "      <td>0.042605</td>\n",
       "      <td>0.006882</td>\n",
       "      <td>0.025660</td>\n",
       "      <td>-0.033505</td>\n",
       "      <td>0.027116</td>\n",
       "      <td>-0.003523</td>\n",
       "      <td>-0.036385</td>\n",
       "      <td>0.071417</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>content_len</th>\n",
       "      <td>0.241384</td>\n",
       "      <td>-0.003337</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>-0.023544</td>\n",
       "      <td>0.308194</td>\n",
       "      <td>0.204101</td>\n",
       "      <td>0.068779</td>\n",
       "      <td>-0.078686</td>\n",
       "      <td>0.246063</td>\n",
       "      <td>0.356481</td>\n",
       "      <td>0.073132</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>comments</th>\n",
       "      <td>-0.147763</td>\n",
       "      <td>0.042605</td>\n",
       "      <td>-0.023544</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.164166</td>\n",
       "      <td>0.290035</td>\n",
       "      <td>0.613961</td>\n",
       "      <td>0.457638</td>\n",
       "      <td>0.349568</td>\n",
       "      <td>0.278942</td>\n",
       "      <td>0.662740</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>favs</th>\n",
       "      <td>0.089313</td>\n",
       "      <td>0.006882</td>\n",
       "      <td>0.308194</td>\n",
       "      <td>0.164166</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.634304</td>\n",
       "      <td>0.416241</td>\n",
       "      <td>0.062877</td>\n",
       "      <td>0.456097</td>\n",
       "      <td>0.587982</td>\n",
       "      <td>0.263239</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>views</th>\n",
       "      <td>0.132385</td>\n",
       "      <td>0.025660</td>\n",
       "      <td>0.204101</td>\n",
       "      <td>0.290035</td>\n",
       "      <td>0.634304</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.396849</td>\n",
       "      <td>0.128654</td>\n",
       "      <td>0.585105</td>\n",
       "      <td>0.406782</td>\n",
       "      <td>0.326427</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>votes_plus</th>\n",
       "      <td>-0.187885</td>\n",
       "      <td>-0.033505</td>\n",
       "      <td>0.068779</td>\n",
       "      <td>0.613961</td>\n",
       "      <td>0.416241</td>\n",
       "      <td>0.396849</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.464168</td>\n",
       "      <td>0.414232</td>\n",
       "      <td>0.449712</td>\n",
       "      <td>0.525081</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>votes_minus</th>\n",
       "      <td>-0.235594</td>\n",
       "      <td>0.027116</td>\n",
       "      <td>-0.078686</td>\n",
       "      <td>0.457638</td>\n",
       "      <td>0.062877</td>\n",
       "      <td>0.128654</td>\n",
       "      <td>0.464168</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.146609</td>\n",
       "      <td>0.088226</td>\n",
       "      <td>0.355458</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>views_lognorm</th>\n",
       "      <td>-0.010376</td>\n",
       "      <td>-0.003523</td>\n",
       "      <td>0.246063</td>\n",
       "      <td>0.349568</td>\n",
       "      <td>0.456097</td>\n",
       "      <td>0.585105</td>\n",
       "      <td>0.414232</td>\n",
       "      <td>0.146609</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.688811</td>\n",
       "      <td>0.546530</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>favs_lognorm</th>\n",
       "      <td>-0.005845</td>\n",
       "      <td>-0.036385</td>\n",
       "      <td>0.356481</td>\n",
       "      <td>0.278942</td>\n",
       "      <td>0.587982</td>\n",
       "      <td>0.406782</td>\n",
       "      <td>0.449712</td>\n",
       "      <td>0.088226</td>\n",
       "      <td>0.688811</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.479476</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>comments_lognorm</th>\n",
       "      <td>-0.001605</td>\n",
       "      <td>0.071417</td>\n",
       "      <td>0.073132</td>\n",
       "      <td>0.662740</td>\n",
       "      <td>0.263239</td>\n",
       "      <td>0.326427</td>\n",
       "      <td>0.525081</td>\n",
       "      <td>0.355458</td>\n",
       "      <td>0.546530</td>\n",
       "      <td>0.479476</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                   post_id   polling  content_len  comments      favs  \\\n",
       "post_id           1.000000  0.081628     0.241384 -0.147763  0.089313   \n",
       "polling           0.081628  1.000000    -0.003337  0.042605  0.006882   \n",
       "content_len       0.241384 -0.003337     1.000000 -0.023544  0.308194   \n",
       "comments         -0.147763  0.042605    -0.023544  1.000000  0.164166   \n",
       "favs              0.089313  0.006882     0.308194  0.164166  1.000000   \n",
       "views             0.132385  0.025660     0.204101  0.290035  0.634304   \n",
       "votes_plus       -0.187885 -0.033505     0.068779  0.613961  0.416241   \n",
       "votes_minus      -0.235594  0.027116    -0.078686  0.457638  0.062877   \n",
       "views_lognorm    -0.010376 -0.003523     0.246063  0.349568  0.456097   \n",
       "favs_lognorm     -0.005845 -0.036385     0.356481  0.278942  0.587982   \n",
       "comments_lognorm -0.001605  0.071417     0.073132  0.662740  0.263239   \n",
       "\n",
       "                     views  votes_plus  votes_minus  views_lognorm  \\\n",
       "post_id           0.132385   -0.187885    -0.235594      -0.010376   \n",
       "polling           0.025660   -0.033505     0.027116      -0.003523   \n",
       "content_len       0.204101    0.068779    -0.078686       0.246063   \n",
       "comments          0.290035    0.613961     0.457638       0.349568   \n",
       "favs              0.634304    0.416241     0.062877       0.456097   \n",
       "views             1.000000    0.396849     0.128654       0.585105   \n",
       "votes_plus        0.396849    1.000000     0.464168       0.414232   \n",
       "votes_minus       0.128654    0.464168     1.000000       0.146609   \n",
       "views_lognorm     0.585105    0.414232     0.146609       1.000000   \n",
       "favs_lognorm      0.406782    0.449712     0.088226       0.688811   \n",
       "comments_lognorm  0.326427    0.525081     0.355458       0.546530   \n",
       "\n",
       "                  favs_lognorm  comments_lognorm  \n",
       "post_id              -0.005845         -0.001605  \n",
       "polling              -0.036385          0.071417  \n",
       "content_len           0.356481          0.073132  \n",
       "comments              0.278942          0.662740  \n",
       "favs                  0.587982          0.263239  \n",
       "views                 0.406782          0.326427  \n",
       "votes_plus            0.449712          0.525081  \n",
       "votes_minus           0.088226          0.355458  \n",
       "views_lognorm         0.688811          0.546530  \n",
       "favs_lognorm          1.000000          0.479476  \n",
       "comments_lognorm      0.479476          1.000000  "
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAm0AAAHdCAYAAABPKU5WAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fd6329189e8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import seaborn as sns\n",
    "corr_matrix = train_df.corr()\n",
    "sns.heatmap(corr_matrix)\n",
    "corr_matrix"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<font color='red'>Вопрос 2.</font> В каком году было больше всего публикаций? (Рассматриваем train_df)\n",
    "- 2014\n",
    "- 2015\n",
    "- 2016\n",
    "- 2017"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2015    23452\n",
      "2016    16316\n",
      "2014    16180\n",
      "2012    15972\n",
      "2013    15537\n",
      "2011    15063\n",
      "2010    13091\n",
      "2009    10783\n",
      "2008     7743\n",
      "Name: year, dtype: int64\n",
      "2015\n"
     ]
    }
   ],
   "source": [
    "train_df['published'] = pd.to_datetime(train_df.published, yearfirst = True)\n",
    "train_df['year'] = [d.year for d in train_df.published]\n",
    "print(train_df.year.value_counts())\n",
    "popyear = train_df.year.value_counts().index[0]\n",
    "print(popyear)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Разбиение на train/valid\n",
    "Используем только признаки 'author', 'flow', 'domain' и 'title'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "features = ['author', 'flow', 'domain','title']\n",
    "train_size = int(0.7 * train_df.shape[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(134137, 93895)"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(train_df),train_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X, y = train_df.ix[:, features],  train_df['favs_lognorm'] #отделяем признаки от целевой переменной\n",
    "\n",
    "X_test = test_df.ix[:, features]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "X_train, X_valid = X.iloc[:train_size, :], X.iloc[train_size:,:]\n",
    "\n",
    "y_train, y_valid = y.iloc[:train_size], y.iloc[train_size:]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Применение TfidfVectorizer\n",
    "\n",
    "**TF-IDF** (от англ. TF — term frequency, IDF — inverse document frequency) — статистическая мера, используемая для оценки важности слова в контексте документа, являющегося частью коллекции документов или корпуса. Вес некоторого слова пропорционален количеству употребления этого слова в документе, и обратно пропорционален частоте употребления слова в других документах коллекции. [Подробнее в источнике](https://ru.wikipedia.org/wiki/TF-IDF)\n",
    "\n",
    "TfidfVectorizer преобразует тексты в матрицу TF-IDF признаков.\n",
    "\n",
    "**Основные параметры TfidfVectorizer в sklearn:**\n",
    "- **min_df** - при построении словаря слова, которые встречаются *реже*, чем указанное значение, игнорируются\n",
    "- **max_df** - при построении словаря слова, которые встречаются *чаще*, чем указанное значение, игнорируются\n",
    "- **analyzer** - определяет, строятся ли признаки по словам или по символам (буквам)\n",
    "- **ngram_range** - определяет, формируются ли признаки только из отдельных слов или из нескольких слов (в случае с analyzer='char' задает количество символов). Например, если указать analyzer='word' и ngram_range=(1,3),то признаки будут формироваться из отдельных слов, из пар слов и из троек слов.\n",
    "- **stop_words** - слова, которые игнорируются при построении матрицы\n",
    "\n",
    "Более подробно с параметрами можно ознакомиться в [документации](http://scikit-learn.org/stable/modules/generated/sklearn.feature_extraction.text.TfidfVectorizer.html)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Инициализируйте TfidfVectorizer с параметрами min_df=3, max_df=0.3 и ngram_range=(1, 3).<br />\n",
    "Примените метод fit_transform к X_train['title'] и метод transform к X_valid['title'] и X_test['title']**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<font color='red'>Вопрос 3.</font> Какой размер у полученного словаря?\n",
    "- 43789\n",
    "- 50624\n",
    "- 93895\n",
    "- 74378"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "vectorizer_title = TfidfVectorizer(min_df=3, max_df=0.3, ngram_range=(1,3))\n",
    "\n",
    "X_train_title = vectorizer_title.fit_transform(X_train['title'])\n",
    "X_valid_title = vectorizer_title.transform(X_valid['title'])\n",
    "X_test_title = vectorizer_title.transform(X_test['title'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'смартфона nokia': 43595,\n",
       " 'версия пакетного менеджера': 15333,\n",
       " 'методологии': 27887,\n",
       " 'tweet': 11176,\n",
       " 'московский номер': 28716,\n",
       " 'на nokia 5800': 29269,\n",
       " 'блога на': 14234,\n",
       " 'oneiric ocelot': 8104,\n",
       " 'обзор apple': 32102,\n",
       " 'школа системного': 49788,\n",
       " 'мобильной разработке': 28339,\n",
       " 'unimrcp voicenavigator': 11315,\n",
       " 'создание чистого css': 44100,\n",
       " 'дизайн привычных': 18641,\n",
       " 'гениальное просто': 17369,\n",
       " 'собственный движок': 43843,\n",
       " 'ni': 7823,\n",
       " 'нуля часть': 31981,\n",
       " 'аккаунта google': 12855,\n",
       " 'аккумуляторов tesla': 12878,\n",
       " 'макеты': 27283,\n",
       " 'кикстартере': 24875,\n",
       " 'почему нельзя': 37675,\n",
       " 'хватит всем': 48530,\n",
       " 'прародитель': 37942,\n",
       " 'нагрузки': 30216,\n",
       " 'nights mobile': 7829,\n",
       " 'vss': 11696,\n",
       " 'rake': 9237,\n",
       " '21 27': 688,\n",
       " 'программируем для psp': 39092,\n",
       " 'радиоволн': 40232,\n",
       " 'moz': 7450,\n",
       " 'wired': 12145,\n",
       " 'программиста часть': 39107,\n",
       " 'простой плагин': 39600,\n",
       " 'debug toolbar': 3291,\n",
       " 'фрилансеров на': 48324,\n",
       " 'роста конверсии сайтов': 41714,\n",
       " 'project anarchy': 8951,\n",
       " 'как вы оцениваете': 24025,\n",
       " '4k intro': 1023,\n",
       " 'для верстки': 19152,\n",
       " 'sencha touch tutorial': 9884,\n",
       " 'общественный': 32663,\n",
       " 'qa': 9092,\n",
       " 'стабильной': 44761,\n",
       " 'прайс': 37899,\n",
       " 'миграции бд для': 27950,\n",
       " 'любителя': 27165,\n",
       " 'определение доминирующих': 33207,\n",
       " 'защищенности': 21786,\n",
       " 'слева': 43400,\n",
       " 'мкс': 28209,\n",
       " 'вместе веселее': 15929,\n",
       " 'прошлого века': 39827,\n",
       " 'из всех': 22348,\n",
       " 'bittorrent dht': 2250,\n",
       " 'азбука n516': 12821,\n",
       " 'на тропу': 30062,\n",
       " 'проектирование веб': 39341,\n",
       " 'егэ по информатике': 20500,\n",
       " 'персонализации': 35068,\n",
       " 'проектов от': 39375,\n",
       " 'конкурента': 25604,\n",
       " 'каким браузером': 24490,\n",
       " 'брагиным': 14471,\n",
       " 'графа': 17831,\n",
       " 'успешный стартап': 47532,\n",
       " 'приема заявок': 38401,\n",
       " 'wizee': 12156,\n",
       " 'sandy': 9751,\n",
       " 'perl воркшоп': 8527,\n",
       " 'keep': 6369,\n",
       " 'газет': 17246,\n",
       " 'уж': 47046,\n",
       " 'соберутся': 43784,\n",
       " 'модель oracle': 28456,\n",
       " 'xe 2013': 12286,\n",
       " 'dribbble': 3652,\n",
       " 'новые проблемы': 31633,\n",
       " 'гаджеты от': 17243,\n",
       " 'delivery': 3318,\n",
       " 'выходит на': 17098,\n",
       " 'от стива': 33813,\n",
       " 'на основе raspberry': 29782,\n",
       " '10 августа': 99,\n",
       " 'презентовать': 38138,\n",
       " 'css рецептов': 3112,\n",
       " 'под hyper': 36169,\n",
       " 'новый мировой рекорд': 31715,\n",
       " 'обвиняет': 32067,\n",
       " 'opera turbo': 8212,\n",
       " 'по крупному': 35830,\n",
       " 'из под': 22415,\n",
       " 'наш взгляд': 30714,\n",
       " 'изображением': 22573,\n",
       " 'slidebar': 10126,\n",
       " 's2': 9654,\n",
       " 'полный экран': 36901,\n",
       " 'сервис вопросов ответов': 42750,\n",
       " 'отправляется': 34202,\n",
       " 'не хотят': 30932,\n",
       " 'как заработать': 24078,\n",
       " 'растения': 40885,\n",
       " 'открывается': 33997,\n",
       " 'pdf': 8485,\n",
       " 'перешли на': 35048,\n",
       " 'июля 2009': 23884,\n",
       " 'по установке': 36000,\n",
       " 'координат': 25921,\n",
       " 'библиотеках': 14060,\n",
       " 'fail': 4137,\n",
       " 'границы': 17820,\n",
       " 'скидок': 43272,\n",
       " 'сервисах из мира': 42790,\n",
       " 'ruby программистов': 9600,\n",
       " 'на маркетинг': 29697,\n",
       " 'новых доменных зон': 31785,\n",
       " 'instead': 5765,\n",
       " 'эффект на': 50341,\n",
       " 'сделано': 42491,\n",
       " 'новый дизайн сайта': 31692,\n",
       " 'на свою': 29972,\n",
       " 'крупные компании': 26412,\n",
       " 'поединок': 36538,\n",
       " 'поставки': 37487,\n",
       " 'комплектация': 25475,\n",
       " 'рекорд по скорости': 41291,\n",
       " 'три одном': 46711,\n",
       " 'бесплатный антивирус': 13964,\n",
       " 'ошибкой': 34442,\n",
       " 'моментов': 28631,\n",
       " 'verizon': 11512,\n",
       " 'неизбежность': 31059,\n",
       " 'google научился': 4961,\n",
       " 'iphone из': 5971,\n",
       " 'сервера помощью': 42710,\n",
       " 'виноват что': 15748,\n",
       " 'forward': 4417,\n",
       " 'hp 3par': 5359,\n",
       " 'открылся для всех': 34019,\n",
       " 'ос на': 33441,\n",
       " 'машинами': 27600,\n",
       " 'формулы': 48197,\n",
       " 'разработке на': 40531,\n",
       " 'шаблонизаторов': 49671,\n",
       " 'заочная': 21465,\n",
       " 'наблюдаем': 30143,\n",
       " 'комитета': 25342,\n",
       " 'голосовое управление': 17659,\n",
       " 'часть заключительная': 49035,\n",
       " 'появятся': 37786,\n",
       " 'флешки от': 48115,\n",
       " 'будильником на': 14611,\n",
       " 'semantic': 9879,\n",
       " 'грозят': 17878,\n",
       " 'выходит продажу': 17100,\n",
       " 'стэнфорд': 45399,\n",
       " 'мода': 28433,\n",
       " 'сайтов по': 42014,\n",
       " 'интеллектуальных': 23178,\n",
       " '125': 297,\n",
       " 'самом': 42079,\n",
       " 'аркады': 13328,\n",
       " 'по разгону': 35923,\n",
       " 'hue': 5465,\n",
       " 'вам не нужен': 14863,\n",
       " 'мультиплатформенный': 28830,\n",
       " 'преимуществ': 38146,\n",
       " 'первой конференции': 34767,\n",
       " 'дыре': 20421,\n",
       " 'учебное': 47718,\n",
       " 'хабр на': 48413,\n",
       " 'этого получилось': 50323,\n",
       " 'курс создание': 26529,\n",
       " 'отследить': 34221,\n",
       " 'фокуса': 48126,\n",
       " 'сети по': 42941,\n",
       " 'культуру': 26493,\n",
       " 'домены на': 20129,\n",
       " 'эриком': 50234,\n",
       " 'столом': 45179,\n",
       " 'как запретить': 24072,\n",
       " 'класса': 24957,\n",
       " 'формат для': 48172,\n",
       " 'платежей на сайте': 35493,\n",
       " 'их решения': 23864,\n",
       " 'по дилетантски или': 35784,\n",
       " 'или карта': 22752,\n",
       " 'touch tutorial': 11073,\n",
       " 'отрезков': 34213,\n",
       " 'gui': 5129,\n",
       " 'создания стартапа': 44131,\n",
       " 'евангелистов или просто': 20453,\n",
       " 'распознавание речи': 40777,\n",
       " 'удобной': 47023,\n",
       " 'создание своей': 44091,\n",
       " 'прицелом': 38811,\n",
       " 'mobile для': 7342,\n",
       " 'читайте': 49416,\n",
       " 'investor day': 5862,\n",
       " 'ru api': 9546,\n",
       " 'добрая': 19869,\n",
       " 'награждение': 30211,\n",
       " 'ли мы': 26805,\n",
       " 'automation': 2010,\n",
       " 'mailto': 6915,\n",
       " 'экран': 49977,\n",
       " 'использовать google': 23634,\n",
       " 'dropbox для': 3676,\n",
       " 'хаоса': 48508,\n",
       " 'файлов средствами': 47866,\n",
       " 'hex': 5280,\n",
       " 'foss news': 4421,\n",
       " 'алгоритмом': 12958,\n",
       " 'microsoft часть': 7240,\n",
       " 'ручка': 41892,\n",
       " 'перспективах': 35091,\n",
       " 'the rope': 10887,\n",
       " 'право ru': 37867,\n",
       " 'выйдет': 16865,\n",
       " 'вёрстке': 17231,\n",
       " 'что вам нужно': 49465,\n",
       " 'хочу работать google': 48707,\n",
       " 'локальный': 27057,\n",
       " 'будет нами': 14590,\n",
       " 'животных': 20756,\n",
       " 'nosql база данных': 7930,\n",
       " 'множественное': 28292,\n",
       " 'objective': 7997,\n",
       " '15 лет назад': 356,\n",
       " 'яндекса для тех': 50578,\n",
       " 'по лицензиям': 35831,\n",
       " 'visual for beginners': 11588,\n",
       " 'дополненная': 20153,\n",
       " 'апдейты': 13268,\n",
       " 'актуальные': 12925,\n",
       " 'парсим': 34624,\n",
       " 'создания сайтов': 44130,\n",
       " 'нии': 31335,\n",
       " 'внешняя': 15980,\n",
       " 'dependency': 3349,\n",
       " 'мессенджера': 27817,\n",
       " 'от xbox': 33716,\n",
       " 'иконки на': 22656,\n",
       " 'строим мультикоптер часть': 45304,\n",
       " 'покупки': 36770,\n",
       " 'социальные': 44401,\n",
       " 'летней давности': 26752,\n",
       " 'котором вы': 26101,\n",
       " 'новая порция': 31425,\n",
       " 'собственный фреймворк на': 43846,\n",
       " 'личности': 26957,\n",
       " 'декабрь январь': 18287,\n",
       " 'из нескольких': 22404,\n",
       " 'планировании': 35407,\n",
       " 'места на': 27823,\n",
       " 'весело': 15382,\n",
       " 'математиков': 27528,\n",
       " 'изобретателей': 22596,\n",
       " 'нулевая': 31972,\n",
       " 'какой то': 24515,\n",
       " 'виды поисковых': 15699,\n",
       " 'открытом виде': 34058,\n",
       " 'google купил': 4953,\n",
       " 'htpc': 5448,\n",
       " 'кибер безопасность еженедельный': 24845,\n",
       " 'искусственного интеллекта': 23516,\n",
       " 'пользуетесь ли вы': 37051,\n",
       " 'приложения инструменты': 38581,\n",
       " 'bender': 2179,\n",
       " '3g 4g': 969,\n",
       " 'game engine': 4594,\n",
       " 'уведомления от': 46887,\n",
       " 'сетевое программирование': 42897,\n",
       " 'родительном падеже': 41593,\n",
       " 'практический курс': 37925,\n",
       " 'организуем': 33397,\n",
       " 'бесплатный вебинар про': 13969,\n",
       " 'app catalog': 1702,\n",
       " 'css без': 3107,\n",
       " 'для центров': 19700,\n",
       " 'стэнфордские': 45402,\n",
       " 'на встречу': 29519,\n",
       " 'обычно': 32765,\n",
       " 'векторной': 15200,\n",
       " 'интеллектуальный': 23177,\n",
       " 'опен сорс': 33130,\n",
       " '77': 1130,\n",
       " 'тот прав записки': 46520,\n",
       " 'di': 3451,\n",
       " 'springsource': 10346,\n",
       " 'конфиг': 25868,\n",
       " 'favicon': 4170,\n",
       " 'youtube для': 12440,\n",
       " 'не 30 строк': 30757,\n",
       " 'panasonic toughbook cf': 8393,\n",
       " 'отладка приложений': 34100,\n",
       " 'бесплатный редактор': 13975,\n",
       " 'usb хаб': 11403,\n",
       " 'под net': 36176,\n",
       " '10 лет': 116,\n",
       " 'поездки': 36543,\n",
       " 'битрикс управление': 14196,\n",
       " 'впечатления': 16295,\n",
       " 'mobile professional': 7330,\n",
       " 'академии': 12845,\n",
       " '2800': 818,\n",
       " 'отношения': 34166,\n",
       " 'учитывается': 47773,\n",
       " 'обновлениями': 32460,\n",
       " 'rescue disk': 9396,\n",
       " 'xhtml': 12309,\n",
       " 'не забывайте': 30803,\n",
       " 'store windows phone': 10522,\n",
       " 'июня 2012': 23900,\n",
       " 'спор': 44540,\n",
       " 'взламывается': 15473,\n",
       " 'rbac': 9272,\n",
       " 'полугодие': 36923,\n",
       " 'времен народов': 16333,\n",
       " 'подобный': 36413,\n",
       " 'google sites': 4884,\n",
       " 'смотри': 43655,\n",
       " 'последний': 37418,\n",
       " 'visual studio online': 11599,\n",
       " 'из twitter': 22324,\n",
       " 'startup weekend': 10467,\n",
       " 'прошивок для android': 39822,\n",
       " 'первой': 34764,\n",
       " 'по шкале ремонтируемости': 36013,\n",
       " 'теперь opensource': 46010,\n",
       " 'прекратит': 38153,\n",
       " 'создания нодового': 44123,\n",
       " 'зачем мне': 21715,\n",
       " 'инжектор': 23015,\n",
       " 'ado net': 1310,\n",
       " 'базе sinatra': 13585,\n",
       " 'мобильный wimax': 28370,\n",
       " 'co working': 2794,\n",
       " 'на wp7': 29402,\n",
       " 'конкурса лучший': 25660,\n",
       " 'одессе': 32824,\n",
       " 'плохие': 35631,\n",
       " 'ввод вывод': 15001,\n",
       " 'никто не видел': 31350,\n",
       " 'твой личный': 45776,\n",
       " 'книг по': 25083,\n",
       " 'приложений игр для': 38527,\n",
       " 'second': 9847,\n",
       " 'включен': 15852,\n",
       " '1970': 458,\n",
       " 'небольшая': 30944,\n",
       " 'разбираемся разработкой windows': 40270,\n",
       " 'домены верхнего': 20125,\n",
       " 'вышла вторая': 17198,\n",
       " 'создания': 44111,\n",
       " 'spamhaus': 10296,\n",
       " 'emc xtremio': 3869,\n",
       " 'мы запускали': 28868,\n",
       " 'слежения за': 43440,\n",
       " 'samsung smarttv': 9721,\n",
       " 'geany': 4635,\n",
       " 'chrome apps': 2628,\n",
       " 'по типу': 35988,\n",
       " 'способа': 44575,\n",
       " 'бег': 13723,\n",
       " 'скоростной': 43315,\n",
       " 'эксперту': 50027,\n",
       " 'продающий дизайн': 39208,\n",
       " 'паролей на': 34601,\n",
       " 'patch': 8445,\n",
       " 'новые функции': 31649,\n",
       " 'быстрее чем': 14734,\n",
       " 'mbr': 7038,\n",
       " 'эксперимент по': 50003,\n",
       " 'miss': 7297,\n",
       " 'книга про': 25091,\n",
       " 'переводов': 34887,\n",
       " 'верхнего': 15367,\n",
       " 'от идеи до': 33742,\n",
       " 'для цифровых': 19702,\n",
       " 'выполнение кода': 16905,\n",
       " 'проекта': 39310,\n",
       " 'высокой емкости': 17030,\n",
       " 'samsung gear': 9713,\n",
       " 'плееров': 35613,\n",
       " 'stp': 10538,\n",
       " 'еле': 20555,\n",
       " 'превращаем': 37952,\n",
       " 'acronis true': 1259,\n",
       " 'верчу': 15374,\n",
       " 'sql server 2014': 10371,\n",
       " 'будет показывать': 14594,\n",
       " 'клавиатуру': 24939,\n",
       " 'оценивает': 34360,\n",
       " 'совещаний': 43912,\n",
       " 'angularjs на': 1614,\n",
       " 'промышленного': 39502,\n",
       " 'слабые': 43390,\n",
       " '29': 821,\n",
       " 'acer iconia': 1243,\n",
       " 'лондонском': 27072,\n",
       " 'популярнее': 37259,\n",
       " 'битву': 14180,\n",
       " 'оптимизации веб': 33249,\n",
       " 'geek week игровой': 4646,\n",
       " 'чурюмова герасименко': 49644,\n",
       " 'yoga': 12415,\n",
       " 'nukem': 7970,\n",
       " 'как почему': 24300,\n",
       " 'цитат': 48892,\n",
       " 'исполнился год': 23560,\n",
       " 'тим бернерс': 46310,\n",
       " 'consumer': 2970,\n",
       " 'не слышали': 30896,\n",
       " 'design challenge': 3361,\n",
       " 'разработчиков на платформе': 40635,\n",
       " 'flexunit': 4337,\n",
       " 'втором': 16641,\n",
       " 'единую': 20518,\n",
       " 'купила': 26502,\n",
       " 'сайт на': 41958,\n",
       " 'робот космонавт': 41552,\n",
       " 'vivotab': 11606,\n",
       " 'дизассемблер': 18684,\n",
       " 'выживание': 16836,\n",
       " 'r2 sp1': 9183,\n",
       " 'на сегодняшний день': 29978,\n",
       " 'продать': 39203,\n",
       " 'секрет': 42568,\n",
       " 'привлечения клиентов': 38327,\n",
       " '12 13': 272,\n",
       " 'системы для': 43184,\n",
       " 'opera поиск по': 8220,\n",
       " 'потенциал': 37563,\n",
       " 'игрового рынка китае': 22119,\n",
       " '22 ноября': 713,\n",
       " 'and drop': 1494,\n",
       " 'lda': 6519,\n",
       " 'официальное': 34311,\n",
       " 'big': 2212,\n",
       " 'двух более': 18221,\n",
       " 'виртуальные путешествия': 15797,\n",
       " 'реальных значений стратегия': 41047,\n",
       " 'все знают': 16386,\n",
       " 'ирландии': 23492,\n",
       " 'вы': 16674,\n",
       " 'windows 95': 11978,\n",
       " 'лебедева': 26641,\n",
       " 'icfpc': 5539,\n",
       " 'рублях': 41764,\n",
       " 'параллельная': 34565,\n",
       " 'операционной системе': 33154,\n",
       " 'ленте': 26706,\n",
       " 'лишние': 26979,\n",
       " 'бросить': 14565,\n",
       " '1700': 418,\n",
       " 'standards': 10436,\n",
       " 'centos часть': 2566,\n",
       " 'доступ данным': 20233,\n",
       " 'talk iphone': 10741,\n",
       " 'my': 7583,\n",
       " 'новые возможности для': 31606,\n",
       " 'для системных': 19591,\n",
       " 'работы со': 40178,\n",
       " 'достояние': 20224,\n",
       " 'намекает': 30374,\n",
       " 'видеообзор ноутбука msi': 15629,\n",
       " 'оригинальный подарок при': 33411,\n",
       " 'предпринимателям': 38042,\n",
       " 'замечание': 21424,\n",
       " 'dell venue': 3330,\n",
       " 'отдыхает': 33911,\n",
       " 'компьютерное': 25522,\n",
       " 'очень полезная': 34389,\n",
       " 'killer': 6405,\n",
       " 'пожертвования': 36557,\n",
       " 'электронной книги pocketbook': 50114,\n",
       " 'симпатичный': 43028,\n",
       " 'пациент': 34710,\n",
       " 'форматы': 48183,\n",
       " 'без ооп': 13771,\n",
       " 'раскрась': 40742,\n",
       " 'под лицензией': 36207,\n",
       " 'из skype': 22321,\n",
       " 'источники данных': 23771,\n",
       " '5d mark': 1068,\n",
       " 'превращение': 37955,\n",
       " '2013 что': 650,\n",
       " 'над дорогой': 30232,\n",
       " 'виртуализацией': 15762,\n",
       " 'видеоаналитики': 15573,\n",
       " 'трафик на': 46612,\n",
       " 'иконок для': 22659,\n",
       " 'куче': 26562,\n",
       " 'федеральных': 47955,\n",
       " 'четвёртая': 49340,\n",
       " 'рукописного': 41814,\n",
       " 'свойства': 42405,\n",
       " 'биндинги': 14151,\n",
       " 'net': 7696,\n",
       " 'сравниваем': 44667,\n",
       " 'на личном опыте': 29678,\n",
       " 'дня': 19764,\n",
       " 'на базе оборудования': 29452,\n",
       " 'грамотная': 17810,\n",
       " 'дайджест интересных': 17973,\n",
       " 'проектирования при разработке': 39361,\n",
       " 'помощью плагина': 37189,\n",
       " 'весь мир': 15400,\n",
       " 'продление': 39242,\n",
       " 'апрельские тезисы': 13293,\n",
       " 'размышления том как': 40420,\n",
       " 'на основе arduino': 29781,\n",
       " 'sony': 10241,\n",
       " 'htc one m8': 5399,\n",
       " 'их на': 23860,\n",
       " 'всеми': 16447,\n",
       " 'flex sdk': 4332,\n",
       " 'репозиториев': 41376,\n",
       " 'google chromecast': 4807,\n",
       " 'ed russia 2011': 3779,\n",
       " 'электронных книг монитора': 50138,\n",
       " 'клиенты': 25038,\n",
       " 'осрв qnx': 33565,\n",
       " 'накопитель от': 30338,\n",
       " 'привлекает': 38323,\n",
       " 'для начинающих менеджеров': 19357,\n",
       " 'выгрузка': 16805,\n",
       " 'встреча microsoft net': 16527,\n",
       " 'начались': 30646,\n",
       " 'для дома': 19207,\n",
       " 'приглашаю на': 38364,\n",
       " 'фотографий на': 48244,\n",
       " 'самых популярных': 42145,\n",
       " 'инсталлятор': 23092,\n",
       " 'частные': 48985,\n",
       " 'каптча': 24611,\n",
       " 'рядом': 41938,\n",
       " 'progress': 8944,\n",
       " 'трёхмерных': 46766,\n",
       " 'software network': 10214,\n",
       " 'наш новый': 30716,\n",
       " 'blackbox': 2273,\n",
       " 'панель для': 34526,\n",
       " 'простейших примеров': 39559,\n",
       " 'капитал': 24602,\n",
       " 'сравнение эффективности': 44663,\n",
       " 'сломать': 43506,\n",
       " 'принимать': 38697,\n",
       " 'работы jquery': 40157,\n",
       " 'прямую трансляцию': 39872,\n",
       " 'локализация': 27042,\n",
       " 'скриптовый язык': 43361,\n",
       " 'версии skype для': 15277,\n",
       " 'parrot ar': 8423,\n",
       " 'на шаг ближе': 30111,\n",
       " 'россию': 41705,\n",
       " 'должна ли': 20002,\n",
       " 'чтения хабра': 49449,\n",
       " 'разработка простого': 40517,\n",
       " 'расширений': 40933,\n",
       " 'тройка': 46727,\n",
       " 'марса от': 27413,\n",
       " 'речи на': 41429,\n",
       " 'тестирование различных': 46147,\n",
       " 'библиотека для работы': 14055,\n",
       " 'кроссбраузерный': 26347,\n",
       " 'день знаний': 18414,\n",
       " 'открытые': 34066,\n",
       " 'вы должны': 16691,\n",
       " 'сотовую сеть': 44326,\n",
       " 'для разработчиков игр': 19525,\n",
       " 'две': 18151,\n",
       " 'задача по': 21252,\n",
       " 'ситуациях': 43207,\n",
       " 'планшетов': 35466,\n",
       " 'robotic': 9472,\n",
       " 'galaxy iii': 4574,\n",
       " 'сеньоритами': 42677,\n",
       " 'прямого': 39862,\n",
       " 'милнера': 28046,\n",
       " 'автоматические': 12626,\n",
       " 'процессоров intel': 39784,\n",
       " 'обращать внимание': 32585,\n",
       " 'стартап': 44884,\n",
       " 'как этим': 24462,\n",
       " 'частоты': 49005,\n",
       " 'казус nokia': 23963,\n",
       " 'погружение': 36145,\n",
       " 'открыла api': 34009,\n",
       " 'способ монетизации': 44554,\n",
       " 'veeam': 11494,\n",
       " 'sslv3': 10420,\n",
       " 'претендентов на': 38214,\n",
       " 'или возвращение': 22680,\n",
       " 'коллективное': 25266,\n",
       " 'условиям': 47484,\n",
       " 'для автоматизации': 19086,\n",
       " 'ui': 11282,\n",
       " 'значков': 21991,\n",
       " 'от skype': 33701,\n",
       " 'core xeon': 3012,\n",
       " 'вакансия': 14842,\n",
       " 'spring': 10338,\n",
       " 'история проекта': 23758,\n",
       " 'приложение на qt': 38478,\n",
       " 'кнопок android': 25129,\n",
       " 'такого докатился': 45708,\n",
       " 'native': 7662,\n",
       " 'храните': 48744,\n",
       " 'googlereader': 5034,\n",
       " 'для парсинга': 19418,\n",
       " 'нужен ли': 31940,\n",
       " 'forum': 4413,\n",
       " 'inspection': 5750,\n",
       " 'переезда на ssd': 34937,\n",
       " 'os для': 8314,\n",
       " 'июня 2011': 23898,\n",
       " 'реверс инжиниринга': 41058,\n",
       " 'на xaml реализуя': 29408,\n",
       " 'сша часть': 45596,\n",
       " 'восьми ферзях': 16273,\n",
       " 's7': 9660,\n",
       " 'thinking functionally': 10903,\n",
       " 'из историй': 22365,\n",
       " 'мозгового': 28589,\n",
       " '10 минут': 126,\n",
       " 'разработки игры под': 40558,\n",
       " 'до ггц': 19800,\n",
       " 'ekb': 3819,\n",
       " 'уравнение': 47427,\n",
       " 'разблокировки': 40277,\n",
       " 'без рук': 13795,\n",
       " 'мне работать': 28235,\n",
       " 'проводят': 38988,\n",
       " 'благо или': 14205,\n",
       " 'слежка': 43441,\n",
       " 'разметку': 40405,\n",
       " 'блог на хабре': 14232,\n",
       " 'app': 1699,\n",
       " 'капризный': 24610,\n",
       " 'не всё': 30783,\n",
       " 'выводов': 16794,\n",
       " 'tor': 11050,\n",
       " 'конструкции': 25703,\n",
       " 'вектор': 15198,\n",
       " 'подручных': 36473,\n",
       " 'components': 2917,\n",
       " 'для науки': 19349,\n",
       " 'сколько': 43280,\n",
       " 'деле': 18352,\n",
       " 'потрясут': 37619,\n",
       " 'секс': 42578,\n",
       " 'ext gwt': 4078,\n",
       " 'на фотографии': 30082,\n",
       " 'пишем драйвер': 35286,\n",
       " '2008 r2 sp1': 541,\n",
       " 'самых маленьких часть': 42142,\n",
       " 'для деплоя': 19198,\n",
       " 'минэкономразвития': 28111,\n",
       " 'водяной': 16065,\n",
       " 'помощью java': 37153,\n",
       " 'чаще': 49103,\n",
       " 'зевс': 21876,\n",
       " 'morpheus': 7406,\n",
       " 'вирусная': 15824,\n",
       " 'поиск на сайте': 36618,\n",
       " 'приобрел': 38756,\n",
       " 'gtk': 5120,\n",
       " 'репортов': 41388,\n",
       " 'использовать': 23633,\n",
       " 'системы распознавания речи': 43193,\n",
       " 'фантастика': 47931,\n",
       " 'hhvm': 5287,\n",
       " 'прогноза': 38993,\n",
       " '78': 1131,\n",
       " 'независимая': 31043,\n",
       " 'javascript без': 6191,\n",
       " 'на python': 29309,\n",
       " 'спецификацию': 44500,\n",
       " 'платформе ios': 35554,\n",
       " 'mashup': 6993,\n",
       " '30 секунд': 869,\n",
       " 'пиццерии': 35272,\n",
       " 'profi gps': 8931,\n",
       " 'использую': 23661,\n",
       " 'школьном': 49808,\n",
       " 'мне не': 28233,\n",
       " 'solo': 10230,\n",
       " 'игра': 22080,\n",
       " 'этого вышло': 50320,\n",
       " 'технологический': 46271,\n",
       " 'создаем свой': 43997,\n",
       " 'бизнес логика': 14092,\n",
       " 'new york': 7791,\n",
       " 'для кого': 19272,\n",
       " 'петафлопс': 35118,\n",
       " 'sharepoint': 9973,\n",
       " 'technical preview': 10803,\n",
       " 'программой': 39146,\n",
       " '2013 28': 628,\n",
       " 'getters setters': 4685,\n",
       " 'поиск vps': 36604,\n",
       " 'событие': 43854,\n",
       " 'rackspace': 9196,\n",
       " 'чтения': 49447,\n",
       " 'samsung анонсировала': 9725,\n",
       " 'треугольники': 46694,\n",
       " 'ненормальное': 31151,\n",
       " 'ubuntu 10 karmic': 11233,\n",
       " 'любую точку': 27191,\n",
       " 'доменного имени': 20098,\n",
       " 'презентации samsung': 38125,\n",
       " 'gigabyte gsmart': 4697,\n",
       " 'межсетевого': 27735,\n",
       " 'двух словах': 18223,\n",
       " 'обнародовала планы': 32381,\n",
       " 'от стива джобса': 33814,\n",
       " 'первое место': 34758,\n",
       " '40 минут': 986,\n",
       " 'информационного': 23432,\n",
       " 'heroes': 5269,\n",
       " 'участвует': 47686,\n",
       " 'php использованием': 8626,\n",
       " 'видеообзор ультрабука планшета': 15643,\n",
       " 'на 90': 29013,\n",
       " 'пол грэм': 36776,\n",
       " 'growth hacking как': 5103,\n",
       " 'водитель дневник': 16051,\n",
       " 'ru на mwc': 9569,\n",
       " 'ms office': 7487,\n",
       " 'или от': 22789,\n",
       " 'расшифровке': 40969,\n",
       " 'прохлада': 39741,\n",
       " 'векторизация': 15199,\n",
       " 'расширениями': 40949,\n",
       " 'линзы': 26863,\n",
       " 'на nancyfx часть': 29257,\n",
       " 'патентов': 34685,\n",
       " 'twister': 11181,\n",
       " '16 декабря': 390,\n",
       " 'начало': 30648,\n",
       " 'мониторинга вашем окружении': 28664,\n",
       " 'индикаторах': 23001,\n",
       " 'сайты для': 42025,\n",
       " 'нитке': 31362,\n",
       " 'лицензию creative commons': 26935,\n",
       " 'прокомментировала': 39477,\n",
       " 'киев бесплатный': 24859,\n",
       " 'streetview': 10551,\n",
       " 'novo': 7953,\n",
       " 'поисковой системы': 36676,\n",
       " 'карту': 24686,\n",
       " 'flash': 4297,\n",
       " 'для': 18811,\n",
       " 'phpinfo': 8657,\n",
       " 'живые обои': 20759,\n",
       " 'eula': 3986,\n",
       " 'unity3dstudent': 11337,\n",
       " 'ted': 10807,\n",
       " 'браузера opera': 14498,\n",
       " 'интерактивное': 23182,\n",
       " 'наконец то вышел': 30333,\n",
       " 'инспектор': 23090,\n",
       " 'антенной': 13226,\n",
       " 'первый планшет на': 34821,\n",
       " 'стать богатым': 45018,\n",
       " 'установок': 47581,\n",
       " 'спешите': 44510,\n",
       " 'текучий': 45843,\n",
       " 'стартапом': 44934,\n",
       " 'номер два': 31813,\n",
       " 'd2': 3191,\n",
       " 'os выйдет': 8313,\n",
       " 'произвольные': 39450,\n",
       " 'пишем кроссбраузерный': 35298,\n",
       " 'финальное': 48053,\n",
       " 'google после': 4991,\n",
       " 'других планетах': 20361,\n",
       " 'правда про google': 37814,\n",
       " 'человечества': 49162,\n",
       " 'пакета для': 34474,\n",
       " 'доклады конференции': 19940,\n",
       " 'удалённого': 46990,\n",
       " 'региональная': 41073,\n",
       " 'деплое': 18448,\n",
       " 'приложений на языке': 38543,\n",
       " 'как мы пишем': 24148,\n",
       " 'управлении': 47358,\n",
       " 'hp mini': 5363,\n",
       " 'интерфейсом': 23404,\n",
       " 'сайта при помощи': 41986,\n",
       " 'cam': 2475,\n",
       " 'радиоастрон': 40231,\n",
       " 'на общественном транспорте': 29759,\n",
       " 'официально': 34303,\n",
       " 'ребята': 41054,\n",
       " 'лекцию': 26686,\n",
       " 'взялись за': 15508,\n",
       " 'оверсан': 32790,\n",
       " 'резервная': 41181,\n",
       " 'released': 9365,\n",
       " 'thinkserver': 10911,\n",
       " 'возврата': 16079,\n",
       " 'яндекс карт': 50534,\n",
       " 'наведении': 30171,\n",
       " 'подход использованию': 36521,\n",
       " 'wolfram language mathematica': 12169,\n",
       " 'вопросов': 16205,\n",
       " 'используйте': 23660,\n",
       " 'про историю': 38865,\n",
       " 'продакшн': 39191,\n",
       " 'ндс для': 30754,\n",
       " 'просили': 39527,\n",
       " 'или поисках': 22804,\n",
       " 'приложения часть': 38596,\n",
       " 'мимики': 28051,\n",
       " 'переключатель раскладки клавиатуры': 34958,\n",
       " 'пишем редактор': 35320,\n",
       " 'google docs появилось': 4823,\n",
       " 'kickstarter': 6393,\n",
       " 'обзор сетевого': 32238,\n",
       " 'чем кажется': 49182,\n",
       " 'steam greenlight': 10483,\n",
       " 'гейтса': 17324,\n",
       " 'для покупок': 19452,\n",
       " 'на пользу': 29835,\n",
       " 'mx2': 7580,\n",
       " '108': 207,\n",
       " 'для разработчиков мобильных': 19526,\n",
       " 'удобством': 47041,\n",
       " '11 50': 226,\n",
       " 'яндекс диске': 50527,\n",
       " 'конкурентом': 25610,\n",
       " 'на колёсах': 29636,\n",
       " 'ищите': 23874,\n",
       " 'mysql performance real': 7604,\n",
       " 'за ошибки': 21009,\n",
       " 'через социальные сети': 49285,\n",
       " 'масштабируемые': 27512,\n",
       " 'корпоративных порталов': 26019,\n",
       " 'идентифицировать': 22239,\n",
       " 'учитывать': 47774,\n",
       " 'не нужны': 30856,\n",
       " 'плагины': 35379,\n",
       " 'маленькую': 27332,\n",
       " 'для mochiweb нагрузкой': 18957,\n",
       " 'новый закон': 31697,\n",
       " 'удобная': 47014,\n",
       " 'программирование как': 39036,\n",
       " 'жителям': 20801,\n",
       " 'цифрового контента': 48903,\n",
       " 'c3': 2444,\n",
       " 'easter': 3752,\n",
       " 'массы': 27484,\n",
       " 'программированию для школьников': 39061,\n",
       " 'принимает': 38696,\n",
       " 'контактам': 25716,\n",
       " 'гайд': 17253,\n",
       " 'товаров на': 46384,\n",
       " 'продуктивнее': 39274,\n",
       " 'базами': 13569,\n",
       " 'поисках': 36644,\n",
       " 'ли все': 26785,\n",
       " 'тепловизор': 46059,\n",
       " 'raid1': 9212,\n",
       " 'опросы': 33238,\n",
       " 'меню на': 27786,\n",
       " 'графический интерфейс': 17863,\n",
       " 'операционные усилители на': 33165,\n",
       " 'бизнес модель': 14094,\n",
       " 'на habrahabr ru': 29158,\n",
       " 'модульной архитектурой': 28510,\n",
       " 'нее': 31039,\n",
       " 'слива': 43451,\n",
       " 'или сам': 22830,\n",
       " 'круглый стол': 26382,\n",
       " 'связи между': 42439,\n",
       " 'веб разработки': 15072,\n",
       " 'автообновление': 12684,\n",
       " 'format': 4405,\n",
       " 'личного кабинета': 26948,\n",
       " 'ресурсов для': 41403,\n",
       " 'об открытых данных': 32047,\n",
       " 'google считает': 5019,\n",
       " 'ua kiev ua': 11214,\n",
       " 'ролика': 41636,\n",
       " 'как увели мои': 24419,\n",
       " 'экстремальный': 50054,\n",
       " 'software configuration': 10206,\n",
       " 'новые тарифы': 31644,\n",
       " 'пожалуйста': 36551,\n",
       " 'jones': 6252,\n",
       " 'патента': 34670,\n",
       " 'wireless': 12146,\n",
       " 'hacker': 5171,\n",
       " 'приглашаю': 38363,\n",
       " 'освободитель': 33453,\n",
       " 'на сервер': 29983,\n",
       " 'parallax scrolling': 8407,\n",
       " 'бэкапа': 14790,\n",
       " 'пожелания': 36554,\n",
       " 'инъекций': 23482,\n",
       " 'методы распознавания': 27904,\n",
       " 'relay': 9360,\n",
       " 'через skype': 49253,\n",
       " 'автоматическая установка': 12624,\n",
       " 'ssh на': 10409,\n",
       " 'общие соображения по': 32671,\n",
       " 'необходимости': 31157,\n",
       " 'маршруты на картах': 27453,\n",
       " 'стартапов часть': 44933,\n",
       " 'заявление': 21801,\n",
       " 'принтер своими руками': 38711,\n",
       " 'iriver': 6016,\n",
       " 'рисками': 41500,\n",
       " 'эластичный': 50059,\n",
       " 'чистка': 49390,\n",
       " 'конструкторов': 25701,\n",
       " 'основы python кратко': 33533,\n",
       " 'java кода': 6157,\n",
       " 'лекций стартап': 26684,\n",
       " 'допустил': 20169,\n",
       " 'по всему': 35763,\n",
       " 'ключевое слово': 25064,\n",
       " 'из этого': 22474,\n",
       " 'новость': 31581,\n",
       " 'термоядерного': 46082,\n",
       " 'ресурсам': 41399,\n",
       " 'полезная утилита': 36791,\n",
       " 'толстой': 46407,\n",
       " 'движущиеся картинки': 18200,\n",
       " 'тиль': 46306,\n",
       " 'никак не': 31337,\n",
       " 'base64': 2118,\n",
       " 'переносим': 34983,\n",
       " 'python часть': 9085,\n",
       " 'желающих': 20693,\n",
       " 'на heroku': 29162,\n",
       " 'microsoft research по': 7180,\n",
       " 'symfony часть': 10688,\n",
       " 'японская': 50590,\n",
       " 'хотя бы': 48689,\n",
       " 'патентный': 34682,\n",
       " 'прецедент': 38223,\n",
       " 'кружки': 26388,\n",
       " 'libgdx': 6597,\n",
       " 'что ты': 49603,\n",
       " 'ms sql': 7491,\n",
       " 'валидатор микроразметки': 14846,\n",
       " 'черных': 49313,\n",
       " 'опыт работы': 33346,\n",
       " 'fireworks': 4276,\n",
       " 'детка': 18515,\n",
       " 'работа веб': 40039,\n",
       " 'генеральный директор': 17334,\n",
       " 'можно узнать': 28580,\n",
       " 'спамерам': 44444,\n",
       " 'валидация': 14848,\n",
       " 'пользовательском': 37033,\n",
       " 'ультрабука sony vaio': 47185,\n",
       " 'black friday': 2262,\n",
       " 'astra': 1939,\n",
       " 'overclocking': 8344,\n",
       " 'власти': 15905,\n",
       " 'телефонов на': 45927,\n",
       " 'lumia 920 820': 6820,\n",
       " 'free lance': 4472,\n",
       " 'функциональные': 48378,\n",
       " 'мониторим': 28652,\n",
       " 'на странице': 30038,\n",
       " 'неправильный': 31194,\n",
       " 'новым': 31771,\n",
       " 'шаблона для joomla': 49665,\n",
       " 'ая встреча': 13541,\n",
       " 'хроники': 48756,\n",
       " 'многофункциональный': 28283,\n",
       " 'альбом': 12988,\n",
       " 'ajax запросов': 1387,\n",
       " 'небольшом': 30959,\n",
       " 'своим': 42350,\n",
       " 'legacy': 6542,\n",
       " 'thunder': 10929,\n",
       " 'far manager': 4157,\n",
       " 'хотите знать': 48681,\n",
       " 'колеса': 25245,\n",
       " 'хотите узнать': 48685,\n",
       " '15 лет': 355,\n",
       " 'убийца': 46864,\n",
       " 'потому': 37602,\n",
       " 'alert': 1404,\n",
       " 'создание полноценного': 44070,\n",
       " 'postgresql часть': 8850,\n",
       " 'новостей игровой индустрии': 31548,\n",
       " 'antivirus': 1637,\n",
       " '1500 рублей': 372,\n",
       " 'он нужен': 33051,\n",
       " ...}"
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Можно посмотреть словарь в виде {'термин': индекс признака,...}\n",
    "vectorizer_title.vocabulary_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "50624"
      ]
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(vectorizer_title.vocabulary_)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<font color='red'>Вопрос 4.</font> Какой индекс у слова 'python'?\n",
    "- 1\n",
    "- 10\n",
    "- 9065\n",
    "- 15679"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "9065"
      ]
     },
     "execution_count": 67,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vectorizer_title.vocabulary_['python']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Инициализируйте TfidfVectorizer, указав analyzer='char'.<br />\n",
    "Примените метод fit_transform к X_train['title'] и метод transform к X_valid['title'] и X_test['title']**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<font color='red'>Вопрос 5.</font> Какой размер у полученного словаря?\n",
    "- 218\n",
    "- 510\n",
    "- 125\n",
    "- 981"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "vectorizer_title_ch = TfidfVectorizer(analyzer='char')\n",
    "\n",
    "X_train_title_ch = vectorizer_title_ch.fit_transform(X_train['title'])\n",
    "X_valid_title_ch = vectorizer_title_ch.transform(X_valid['title'])\n",
    "X_test_title_ch = vectorizer_title_ch.transform(X_test['title'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{' ': 0,\n",
       " '!': 1,\n",
       " '\"': 2,\n",
       " '#': 3,\n",
       " '$': 4,\n",
       " '%': 5,\n",
       " '&': 6,\n",
       " \"'\": 7,\n",
       " '(': 8,\n",
       " ')': 9,\n",
       " '*': 10,\n",
       " '+': 11,\n",
       " ',': 12,\n",
       " '-': 13,\n",
       " '.': 14,\n",
       " '/': 15,\n",
       " '0': 16,\n",
       " '1': 17,\n",
       " '2': 18,\n",
       " '3': 19,\n",
       " '4': 20,\n",
       " '5': 21,\n",
       " '6': 22,\n",
       " '7': 23,\n",
       " '8': 24,\n",
       " '9': 25,\n",
       " ':': 26,\n",
       " ';': 27,\n",
       " '<': 28,\n",
       " '=': 29,\n",
       " '>': 30,\n",
       " '?': 31,\n",
       " '@': 32,\n",
       " '[': 33,\n",
       " '\\\\': 34,\n",
       " ']': 35,\n",
       " '^': 36,\n",
       " '_': 37,\n",
       " '`': 38,\n",
       " 'a': 39,\n",
       " 'b': 40,\n",
       " 'c': 41,\n",
       " 'd': 42,\n",
       " 'e': 43,\n",
       " 'f': 44,\n",
       " 'g': 45,\n",
       " 'h': 46,\n",
       " 'i': 47,\n",
       " 'j': 48,\n",
       " 'k': 49,\n",
       " 'l': 50,\n",
       " 'm': 51,\n",
       " 'n': 52,\n",
       " 'o': 53,\n",
       " 'p': 54,\n",
       " 'q': 55,\n",
       " 'r': 56,\n",
       " 's': 57,\n",
       " 't': 58,\n",
       " 'u': 59,\n",
       " 'v': 60,\n",
       " 'w': 61,\n",
       " 'x': 62,\n",
       " 'y': 63,\n",
       " 'z': 64,\n",
       " '{': 65,\n",
       " '|': 66,\n",
       " '}': 67,\n",
       " '~': 68,\n",
       " '\\x92': 69,\n",
       " '\\x97': 70,\n",
       " '\\xa0': 71,\n",
       " '£': 72,\n",
       " '¥': 73,\n",
       " '§': 74,\n",
       " '©': 75,\n",
       " '«': 76,\n",
       " '\\xad': 77,\n",
       " '®': 78,\n",
       " '°': 79,\n",
       " '±': 80,\n",
       " '²': 81,\n",
       " 'µ': 82,\n",
       " '·': 83,\n",
       " '¹': 84,\n",
       " '»': 85,\n",
       " '¼': 86,\n",
       " '½': 87,\n",
       " '×': 88,\n",
       " 'à': 89,\n",
       " 'á': 90,\n",
       " 'â': 91,\n",
       " 'ä': 92,\n",
       " 'ç': 93,\n",
       " 'é': 94,\n",
       " 'ë': 95,\n",
       " 'ï': 96,\n",
       " 'ñ': 97,\n",
       " 'ò': 98,\n",
       " 'ó': 99,\n",
       " 'ö': 100,\n",
       " 'ø': 101,\n",
       " 'ù': 102,\n",
       " 'ú': 103,\n",
       " 'ü': 104,\n",
       " 'ı': 105,\n",
       " 'ǝ': 106,\n",
       " 'ɟ': 107,\n",
       " 'ʇ': 108,\n",
       " '́': 109,\n",
       " '̆': 110,\n",
       " '̱': 111,\n",
       " '̲': 112,\n",
       " '̶': 113,\n",
       " 'β': 114,\n",
       " 'δ': 115,\n",
       " 'λ': 116,\n",
       " 'μ': 117,\n",
       " 'π': 118,\n",
       " 'ω': 119,\n",
       " 'а': 120,\n",
       " 'б': 121,\n",
       " 'в': 122,\n",
       " 'г': 123,\n",
       " 'д': 124,\n",
       " 'е': 125,\n",
       " 'ж': 126,\n",
       " 'з': 127,\n",
       " 'и': 128,\n",
       " 'й': 129,\n",
       " 'к': 130,\n",
       " 'л': 131,\n",
       " 'м': 132,\n",
       " 'н': 133,\n",
       " 'о': 134,\n",
       " 'п': 135,\n",
       " 'р': 136,\n",
       " 'с': 137,\n",
       " 'т': 138,\n",
       " 'у': 139,\n",
       " 'ф': 140,\n",
       " 'х': 141,\n",
       " 'ц': 142,\n",
       " 'ч': 143,\n",
       " 'ш': 144,\n",
       " 'щ': 145,\n",
       " 'ъ': 146,\n",
       " 'ы': 147,\n",
       " 'ь': 148,\n",
       " 'э': 149,\n",
       " 'ю': 150,\n",
       " 'я': 151,\n",
       " 'ё': 152,\n",
       " 'і': 153,\n",
       " 'ї': 154,\n",
       " 'ј': 155,\n",
       " 'ў': 156,\n",
       " 'ѣ': 157,\n",
       " 'қ': 158,\n",
       " '۞': 159,\n",
       " '۩': 160,\n",
       " '฿': 161,\n",
       " '\\u2002': 162,\n",
       " '\\u2006': 163,\n",
       " '\\u2009': 164,\n",
       " '\\u200b': 165,\n",
       " '\\u200e': 166,\n",
       " '‐': 167,\n",
       " '‑': 168,\n",
       " '–': 169,\n",
       " '—': 170,\n",
       " '―': 171,\n",
       " '‘': 172,\n",
       " '’': 173,\n",
       " '“': 174,\n",
       " '”': 175,\n",
       " '„': 176,\n",
       " '•': 177,\n",
       " '…': 178,\n",
       " '\\u2028': 179,\n",
       " '\\u202c': 180,\n",
       " '\\u202d': 181,\n",
       " '′': 182,\n",
       " '″': 183,\n",
       " '‽': 184,\n",
       " '⁰': 185,\n",
       " '€': 186,\n",
       " '№': 187,\n",
       " '™': 188,\n",
       " '⅓': 189,\n",
       " '⅜': 190,\n",
       " '←': 191,\n",
       " '↑': 192,\n",
       " '→': 193,\n",
       " '↔': 194,\n",
       " '↺': 195,\n",
       " '⇄': 196,\n",
       " '⇒': 197,\n",
       " '⇨': 198,\n",
       " '−': 199,\n",
       " '∙': 200,\n",
       " '≈': 201,\n",
       " '≠': 202,\n",
       " '━': 203,\n",
       " '▼': 204,\n",
       " '☠': 205,\n",
       " '☢': 206,\n",
       " '☣': 207,\n",
       " '☼': 208,\n",
       " '♠': 209,\n",
       " '♥': 210,\n",
       " '♫': 211,\n",
       " '❤': 212,\n",
       " '印': 213,\n",
       " '打': 214,\n",
       " '机': 215,\n",
       " '\\ufeff': 216,\n",
       " '！': 217}"
      ]
     },
     "execution_count": 69,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Здесь так же можно посмотреть словарь\n",
    "#Заметьте насколько отличаются словари для TfidfVectorizer с analyzer='word' и analyzer='char'\n",
    "vectorizer_title_ch.vocabulary_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "218"
      ]
     },
     "execution_count": 70,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(vectorizer_title_ch.vocabulary_)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Работа с категориальными признаками\n",
    "\n",
    "Для обработки категориальных признаков 'author', 'flow', 'domain' мы будем использовать DictVectorizer из sklearn."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>author</th>\n",
       "      <th>flow</th>\n",
       "      <th>domain</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>@Tapac</td>\n",
       "      <td>develop</td>\n",
       "      <td>habrahabr.ru</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>@DezmASter</td>\n",
       "      <td>design</td>\n",
       "      <td>habrahabr.ru</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>@DezmASter</td>\n",
       "      <td>design</td>\n",
       "      <td>habrahabr.ru</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>@Taoorus</td>\n",
       "      <td>design</td>\n",
       "      <td>habrahabr.ru</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>@dennydo</td>\n",
       "      <td>NaN</td>\n",
       "      <td>geektimes.ru</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       author     flow        domain\n",
       "0      @Tapac  develop  habrahabr.ru\n",
       "1  @DezmASter   design  habrahabr.ru\n",
       "2  @DezmASter   design  habrahabr.ru\n",
       "3    @Taoorus   design  habrahabr.ru\n",
       "4    @dennydo      NaN  geektimes.ru"
      ]
     },
     "execution_count": 71,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "feats = ['author', 'flow', 'domain']\n",
    "X_train[feats][:5]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Рассмотрим как он работает на примере первых пяти строк"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>author</th>\n",
       "      <th>flow</th>\n",
       "      <th>domain</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>@Tapac</td>\n",
       "      <td>develop</td>\n",
       "      <td>habrahabr.ru</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>@DezmASter</td>\n",
       "      <td>design</td>\n",
       "      <td>habrahabr.ru</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>@DezmASter</td>\n",
       "      <td>design</td>\n",
       "      <td>habrahabr.ru</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>@Taoorus</td>\n",
       "      <td>design</td>\n",
       "      <td>habrahabr.ru</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>@dennydo</td>\n",
       "      <td>-</td>\n",
       "      <td>geektimes.ru</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       author     flow        domain\n",
       "0      @Tapac  develop  habrahabr.ru\n",
       "1  @DezmASter   design  habrahabr.ru\n",
       "2  @DezmASter   design  habrahabr.ru\n",
       "3    @Taoorus   design  habrahabr.ru\n",
       "4    @dennydo        -  geektimes.ru"
      ]
     },
     "execution_count": 72,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#сначала заполняем пропуски прочерком\n",
    "X_train[feats][:5].fillna('-')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{0: {'author': '@Tapac', 'domain': 'habrahabr.ru', 'flow': 'develop'},\n",
       " 1: {'author': '@DezmASter', 'domain': 'habrahabr.ru', 'flow': 'design'},\n",
       " 2: {'author': '@DezmASter', 'domain': 'habrahabr.ru', 'flow': 'design'},\n",
       " 3: {'author': '@Taoorus', 'domain': 'habrahabr.ru', 'flow': 'design'},\n",
       " 4: {'author': '@dennydo', 'domain': 'geektimes.ru', 'flow': '-'}}"
      ]
     },
     "execution_count": 73,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Преобразуем датафрейм в словарь, где ключами являются индексы объектов (именно для этого мы транспонировали датафрейм),\n",
    "#а значениями являются словари в виде 'название_колонки':'значение'\n",
    "X_train[feats][:5].fillna('-').T.to_dict()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dict_values([{'domain': 'habrahabr.ru', 'flow': 'develop', 'author': '@Tapac'}, {'domain': 'habrahabr.ru', 'flow': 'design', 'author': '@DezmASter'}, {'domain': 'habrahabr.ru', 'flow': 'design', 'author': '@DezmASter'}, {'domain': 'habrahabr.ru', 'flow': 'design', 'author': '@Taoorus'}, {'domain': 'geektimes.ru', 'flow': '-', 'author': '@dennydo'}])"
      ]
     },
     "execution_count": 74,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#В DictVectorizer нам нужно будет передать список словарей для каждого объекта в виде 'название_колонки':'значение',\n",
    "#поэтому используем .values()\n",
    "X_train[feats][:5].fillna('-').T.to_dict().values()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<5x9 sparse matrix of type '<class 'numpy.float64'>'\n",
       "\twith 15 stored elements in Compressed Sparse Row format>"
      ]
     },
     "execution_count": 75,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#В итоге получается разреженная матрица\n",
    "dict_vect = DictVectorizer()\n",
    "dict_vect_matrix = dict_vect.fit_transform(X_train[feats][:5].fillna('-').T.to_dict().values())\n",
    "dict_vect_matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 0.,  0.,  1.,  0.,  0.,  1.,  0.,  0.,  1.],\n",
       "       [ 1.,  0.,  0.,  0.,  0.,  1.,  0.,  1.,  0.],\n",
       "       [ 1.,  0.,  0.,  0.,  0.,  1.,  0.,  1.,  0.],\n",
       "       [ 0.,  1.,  0.,  0.,  0.,  1.,  0.,  1.,  0.],\n",
       "       [ 0.,  0.,  0.,  1.,  1.,  0.,  1.,  0.,  0.]])"
      ]
     },
     "execution_count": 76,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Но можно преобразовать ее в numpy array с помощью .toarray()\n",
    "dict_vect_matrix.toarray()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(5, 9)"
      ]
     },
     "execution_count": 77,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#В получившейся матрице 5 строк (по числу объектов) и 9 столбцов\n",
    "#Далее разберемся почему колонок именно 9\n",
    "dict_vect_matrix.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Посмотрим сколько уникальных значений в каждой колонке.<br />\n",
    "Суммарно их 9 - столько же, сколько и колонок. Это объясняется тем, что для категориальных признаков со строковыми значениями DictVectorizer делает кодирование бинарными признаками - каждому уникальному значению признака соответствует один новый бинарный признак, который равен 1 только в том случае, если в исходной матрице этот признак принимает значение, которому соответствует эта колонка новой матрицы."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "author 4\n",
      "flow 3\n",
      "domain 2\n"
     ]
    }
   ],
   "source": [
    "for col in feats:\n",
    "    print(col,len(X_train[col][:5].fillna('-').unique()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Также можно посмотреть что означает каждая колонка полученной матрицы"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['author=@DezmASter',\n",
       " 'author=@Taoorus',\n",
       " 'author=@Tapac',\n",
       " 'author=@dennydo',\n",
       " 'domain=geektimes.ru',\n",
       " 'domain=habrahabr.ru',\n",
       " 'flow=-',\n",
       " 'flow=design',\n",
       " 'flow=develop']"
      ]
     },
     "execution_count": 79,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#например, самая первая колонка называется 'author=@DezmASter' - то есть принимает значение 1 только если автор @DezmASter\n",
    "dict_vect.feature_names_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Инициализируйте DictVectorizer с параметрами по умолчанию.<br />\n",
    "Примените метод fit_transform к 'author', 'flow', 'domain' X_train и метод transform к X_valid и X_test**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "vectorizer_feats = DictVectorizer()\n",
    "\n",
    "X_train_feats = vectorizer_feats.fit_transform(X_train[feats].fillna('-').T.to_dict().values())\n",
    "X_valid_feats = vectorizer_feats.transform(X_valid[feats].fillna('-').T.to_dict().values())\n",
    "X_test_feats = vectorizer_feats.transform(X_test[feats].fillna('-').T.to_dict().values())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(93895, 17869)"
      ]
     },
     "execution_count": 86,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train_feats.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Соединим все полученные матрицы при помощи scipy.sparse.hstack()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "X_train_new = scipy.sparse.hstack([X_train_title, X_train_feats, X_train_title_ch])\n",
    "X_valid_new = scipy.sparse.hstack([X_valid_title, X_valid_feats, X_valid_title_ch])\n",
    "X_test_new =  scipy.sparse.hstack([X_test_title, X_test_feats, X_test_title_ch])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Обучение модели\n",
    "\n",
    "Далее будем использовать Ridge, линейную модель с l2-регуляризацией.\n",
    "[Документация](http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.Ridge.html)\n",
    "\n",
    "Основной параметр Ridge - **alpha, коэффициент регуляризации**. Регуляризация используется для улучшения обобщающей способности модели - прибавляя к функционалу потерь сумму квадратов весов, умноженную на коэффициент регуляризации (та самая alpha), мы штрафуем модель за слишком большие значения весов и не позволяем ей переобучаться. Чем больше этот коээфициент, тем сильнее эффект."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Обучите две модели на X_train_new, y_train, задав в первой alpha=0.1 и random_state = 1, а во второй alpha=1.0 и random_state = 1**\n",
    "\n",
    "**Рассчитайте среднеквадратичную ошибку каждой модели (mean_squared_error). Сравните значения ошибки на обучающей и тестовой выборках и ответьте на вопросы.**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<font color='red'>Вопрос 6.</font> Выберите верные утверждения:\n",
    "- обе модели показывают одинаковый результат (среднеквадратичная ошибка отличается не больше чем на тысячные), регуляризация ничего не меняет\n",
    "- при alpha=0.1 модель переобучается\n",
    "- среднеквадратичная ошибка первой модели на тесте меньше\n",
    "- при alpha=1.0 у модели обощающая способность лучше, чем у при alpha=0.1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 11.9 s, sys: 48 ms, total: 12 s\n",
      "Wall time: 12.1 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "model1 = Ridge(alpha = 0.1, random_state = 1)\n",
    "model1.fit(X_train_new, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Ошибка на трейне 0.184107031582\n",
      "Ошибка на тесте 0.996233534497\n"
     ]
    }
   ],
   "source": [
    "train_preds1 = model1.predict(X_train_new)\n",
    "valid_preds1 = model1.predict(X_valid_new)\n",
    "\n",
    "print('Ошибка на трейне',mean_squared_error(y_train, train_preds1))\n",
    "print('Ошибка на тесте',mean_squared_error(y_valid, valid_preds1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 4.6 s, sys: 4 ms, total: 4.6 s\n",
      "Wall time: 4.61 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "model2 = Ridge(alpha = 1, random_state = 1)\n",
    "model2.fit(X_train_new, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Ошибка на трейне 0.28624407471\n",
      "Ошибка на тесте 0.80695156251\n"
     ]
    }
   ],
   "source": [
    "train_preds2 = model2.predict(X_train_new)\n",
    "valid_preds2 = model2.predict(X_valid_new)\n",
    "\n",
    "print('Ошибка на трейне',mean_squared_error(y_train, train_preds2))\n",
    "print('Ошибка на тесте',mean_squared_error(y_valid, valid_preds2))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Baseline\n",
    "\n",
    "**Теперь попытаемся получить бейзлайн для соревования - используйте Ridge с параметрами по умолчанию и обучите модель на всех данных - соедините X_train_new X_valid_new (используйте scipy.sparse.vstack()), а целевой переменной будет y.**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 8.36 s, sys: 48 ms, total: 8.41 s\n",
      "Wall time: 8.4 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "model = Ridge()\n",
    "\n",
    "X_train_all = scipy.sparse.vstack([X_train_new, X_valid_new])\n",
    "model.fit(X_train_all, y)\n",
    "\n",
    "test_preds = model.predict(X_test_new)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "sample_submission = pd.read_csv('sample_submission.csv', index_col='url')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>favs_lognorm</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>url</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>https://habrahabr.ru/post/314080/</th>\n",
       "      <td>0.323588</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>https://habrahabr.ru/company/plesk/blog/313732/</th>\n",
       "      <td>0.560105</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>https://habrahabr.ru/company/etagi/blog/314000/</th>\n",
       "      <td>0.667039</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>https://habrahabr.ru/company/knopka/blog/314030/</th>\n",
       "      <td>0.149563</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>https://geektimes.ru/company/audiomania/blog/282058/</th>\n",
       "      <td>0.670754</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                    favs_lognorm\n",
       "url                                                             \n",
       "https://habrahabr.ru/post/314080/                       0.323588\n",
       "https://habrahabr.ru/company/plesk/blog/313732/         0.560105\n",
       "https://habrahabr.ru/company/etagi/blog/314000/         0.667039\n",
       "https://habrahabr.ru/company/knopka/blog/314030/        0.149563\n",
       "https://geektimes.ru/company/audiomania/blog/28...      0.670754"
      ]
     },
     "execution_count": 100,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sample_submission.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ridge_submission = sample_submission.copy()\n",
    "ridge_submission['favs_lognorm'] = test_preds\n",
    "ridge_submission.to_csv('ridge_baseline.csv') #а вот и бейзлайн"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}