models_rf_and_dtree_withGS.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "from sklearn.base import BaseEstimator, TransformerMixin\n",
    "from sklearn.preprocessing import OneHotEncoder, StandardScaler\n",
    "from sklearn.impute import SimpleImputer\n",
    "from sklearn.pipeline import FeatureUnion, Pipeline\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn import metrics\n",
    "\n",
    "import warnings\n",
    "warnings.simplefilter('ignore')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "class FeatureSelector(BaseEstimator, TransformerMixin):\n",
    "    \"\"\"\n",
    "    The class provides basic functionality for retrieving\n",
    "    a subset of columns from the dataset.\n",
    "    \"\"\"\n",
    "    \n",
    "    def __init__(self, feature_names):\n",
    "        \"\"\"\n",
    "        Initialize class instance by setting\n",
    "        a list of columns to retrieve from the dataset.\n",
    "        \"\"\"\n",
    "        BaseEstimator.__init__(self)\n",
    "        TransformerMixin.__init__(self)\n",
    "        self.feature_names = feature_names\n",
    "        \n",
    "    def fit(self, X, y=None):\n",
    "        \"\"\"\n",
    "        Fit FeatureSelector to X, but really do nothing.\n",
    "        Return self.\n",
    "        \"\"\"\n",
    "        return self\n",
    "    \n",
    "    def transform(self, X, y=None):\n",
    "        \"\"\"\n",
    "        Transform X using feature selection. \n",
    "        Return column-subset of X.\n",
    "        \"\"\"\n",
    "        return X[self.feature_names]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "class FeatureGenerator(BaseEstimator, TransformerMixin):\n",
    "    \"\"\"\n",
    "    Required columns: table. X is DataFrame.\n",
    "    \"\"\"\n",
    "    \n",
    "    features = {\n",
    "             'ABS (антиблокировочная система)',\n",
    "             'AUX/iPod',\n",
    "             'Bluetooth',\n",
    "             'CD/MP3 проигрыватель',\n",
    "             'ESP (система поддержания динамической стабильности)',\n",
    "             'USB',\n",
    "             'Автозапуск двигателя',\n",
    "             'Антипробуксовочная система',\n",
    "             'Датчик дождя',\n",
    "             'Иммобилайзер',\n",
    "             'Камера заднего вида',\n",
    "             'Климат-контроль',\n",
    "             'Кондиционер',\n",
    "             'Контроль мертвых зон на зеркалах',\n",
    "             'Круиз-контроль',\n",
    "             'Ксеноновые фары',\n",
    "             'Легкосплавные диски',\n",
    "             'Люк',\n",
    "             'Материал салона - натуральная кожа',\n",
    "             'Мультимедийный экран',\n",
    "             'Обогрев зеркал',\n",
    "             'Обогрев лобового стекла',\n",
    "             'Обогрев руля',\n",
    "             'Обогрев сидений',\n",
    "             'Панорамная крыша',\n",
    "             'Парктроники',\n",
    "             'Подушки безопасности боковые',\n",
    "             'Подушки безопасности задние',\n",
    "             'Подушки безопасности передние',\n",
    "             'Противотуманные фары',\n",
    "             'Рейлинги на крыше',\n",
    "             'Светодиодные фары',\n",
    "             'Сигнализация',\n",
    "             'Системы помощи',\n",
    "             'Управление мультимедиа с руля',\n",
    "             'Фаркоп',\n",
    "             'Штатная навигация',\n",
    "             'Электрорегулировка сидений',\n",
    "             'Электростеклоподъемники задние',\n",
    "             'Электростеклоподъемники передние'}\n",
    "    \n",
    "    def fit(self, X, y=None):\n",
    "        return self\n",
    "    \n",
    "    def transform(self, X, y=None):\n",
    "        X = X.copy()\n",
    "        self.add_features(X)\n",
    "        return X[self.features].values\n",
    "    \n",
    "    def add_features(self, X):\n",
    "        for imp in self.features:\n",
    "            X[imp] = X.table.apply(lambda x: int(imp in x)).astype('int8')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "class ToIntTransformer(BaseEstimator, TransformerMixin):\n",
    "    \"\"\"\n",
    "    Required columns: volume, show, run, pages, update, year.\n",
    "    \"\"\"\n",
    "    \n",
    "    def fit(self, X, y=None):\n",
    "        return self\n",
    "    \n",
    "    def transform(self, X, y=None):\n",
    "        X = X.copy()\n",
    "        # self.to_int(X, 'cost', ' ')\n",
    "        self.to_int(X, 'volume', ' см3')\n",
    "        self.fix_show(X)\n",
    "        self.fix_run(X)\n",
    "        self.fix_dates(X)\n",
    "        self.add_restyle(X)\n",
    "        self.add_upd_flags(X)\n",
    "        self.year_to_old(X)\n",
    "        return X.drop(['model', 'update', 'pages', 'year'], axis=1).values\n",
    "    \n",
    "    def to_int(self, X, column_name, phrase):\n",
    "        X[column_name] = X[column_name].str.replace(phrase, '').astype('int32')\n",
    "        \n",
    "    def year_to_old(self, X):\n",
    "        X['age'] = (2019 - X['year']).astype('int8')\n",
    "        \n",
    "    def fix_show(self, X):\n",
    "        X['today_views'] = X['show'].str.extract('\\+(.+) ')\n",
    "        X['show'] = X['show'].str.extract('(.*)' + ' '*25)\n",
    "        X.rename(columns={'show': 'all_views'}, inplace=True)\n",
    "        today_view_mask = pd.isna(X['today_views'])\n",
    "        X.loc[today_view_mask, 'today_views'] = X[today_view_mask]['all_views']\n",
    "        X['today_views'] = X['today_views'].astype('int')\n",
    "        X['all_views'] = X['all_views'].astype('int')\n",
    "        \n",
    "    def fix_run(self, X):\n",
    "        X['run'] = X['run'].str.replace(' км', '')\n",
    "        miles_mask = X['run'].str.endswith(' миль')\n",
    "        X.loc[miles_mask, 'run'] = X[miles_mask]['run'].str.replace(' миль', '').astype('int') * 1.60934\n",
    "        X['run'] = X['run'].astype('int')\n",
    "        \n",
    "    def add_restyle(self, X):\n",
    "        self.create_model(X)\n",
    "        X['is_restyle'] = X['model'].str.endswith('(рестайлинг)').astype('int8')\n",
    "        \n",
    "    def add_upd_flags(self, X):\n",
    "        X['modified'] = X['update'].apply(\n",
    "            lambda x: int(not x.split()[0] == 'Опубликовано')).astype('int8')\n",
    "        X['up'] = X['update'].apply(\n",
    "            lambda x: int(len(x.split()) == 4)).astype('int8')\n",
    "    \n",
    "    def create_model(self, X):\n",
    "        two_word_names = ('Alfa Romeo', 'Great Wall', 'Lada (ВАЗ)')\n",
    "        two_word_names_mask = X['pages'].str.startswith(two_word_names)\n",
    "        X.loc[two_word_names_mask, 'model'] = (X[two_word_names_mask]['pages'].str.split()\n",
    "                                                .apply(lambda name: ' '.join(name[2:])))\n",
    "        X.loc[~two_word_names_mask, 'model'] = (X[~two_word_names_mask]['pages'].str.split()\n",
    "                                                 .apply(lambda name: ' '.join(name[1:])))\n",
    "    \n",
    "    def fix_dates(self, df):\n",
    "        today=pd.Timestamp(2019, 11, 23)\n",
    "        df['days_ago'] = df['update'].apply(\n",
    "            lambda x: (today - pd.Timestamp(x.split()[1])).days)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "class CatTransformer(BaseEstimator, TransformerMixin):\n",
    "    \"\"\"\n",
    "    Required columns: cuzov, fuel, pages, region, update.\n",
    "    \"\"\"    \n",
    "    \n",
    "    def fit(self, X, y=None):\n",
    "        return self\n",
    "    \n",
    "    def transform(self, X, y=None):\n",
    "        X = X.copy()\n",
    "        self.cut(X, ['cuzov', 'fuel'])\n",
    "        self.fix_names(X)\n",
    "        self.fix_region(X)\n",
    "        return X.drop(['update'], axis=1)\n",
    "    \n",
    "    def cut(self, X, column_names):\n",
    "        for col in column_names:\n",
    "            X[col] = X[col].apply(lambda x: x.split()[0])\n",
    "        \n",
    "    def fix_names(self, df):\n",
    "        two_word_names = ('Alfa Romeo', 'Great Wall', 'Lada (ВАЗ)')\n",
    "        two_word_names_mask = df['pages'].str.startswith(two_word_names)\n",
    "        # df.loc[two_word_names_mask, 'model'] = (df[two_word_names_mask]['pages'].str.split()\n",
    "        #                                         .apply(lambda name: ' '.join(name[2:])))\n",
    "        # df.loc[~two_word_names_mask, 'model'] = (df[~two_word_names_mask]['pages'].str.split()\n",
    "        #                                          .apply(lambda name: ' '.join(name[1:])))\n",
    "        df.loc[two_word_names_mask, 'pages'] = (df[two_word_names_mask]['pages'].str.split()\n",
    "                                                .apply(lambda name: ' '.join(name[:2])))\n",
    "        df.loc[~two_word_names_mask, 'pages'] = (df[~two_word_names_mask]['pages'].str.split()\n",
    "                                                 .apply(lambda name: ' '.join(name[:1])))\n",
    "        df.rename(columns={'pages': 'brand'}, inplace=True)\n",
    "        vc = df.brand.value_counts()\n",
    "        df.brand = df.brand.apply(lambda x: x if vc[x] > 100 else 'other')\n",
    "        \n",
    "    def fix_region(self, X):\n",
    "        \n",
    "        def _get_region(lst):\n",
    "            if len(lst) == 1:\n",
    "                return lst[0]\n",
    "            return lst[1]\n",
    "        \n",
    "        tmp = list(map(lambda s: s.split(', '), X.region))\n",
    "        X.region = list(map(_get_region, tmp))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "class ColumnTranslation:\n",
    "    \n",
    "    def __init__(self, column_name, to_save, default='Other'):\n",
    "        self.column_name = column_name\n",
    "        self.to_save = to_save\n",
    "        self.default = default\n",
    "\n",
    "\n",
    "class Translator(BaseEstimator, TransformerMixin):\n",
    "    \n",
    "    def __init__(self, translations):\n",
    "        self.translations = translations[:]\n",
    "        \n",
    "    def fit(self, X, y=None):\n",
    "        return self\n",
    "    \n",
    "    def transform(self, X, y=None):\n",
    "        X = X.copy()\n",
    "        for tr in self.translations:\n",
    "            X[tr.column_name] = X[tr.column_name].apply(\n",
    "                lambda x: x if x in tr.to_save else tr.default)\n",
    "        return X[[tr.column_name for tr in self.translations]].values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "avto = pd.read_csv('final.csv')\n",
    "\n",
    "X = avto.drop(['cost'], axis=1)\n",
    "y = avto.cost.apply(lambda x: int(x.replace(' ', '')))\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y ,test_size=0.2, random_state=42)\n",
    "\n",
    "feat_pipeline = Pipeline(\n",
    "    steps=[\n",
    "        ('feat_selector', FeatureSelector(['table'])),\n",
    "        ('feat_generator', FeatureGenerator())\n",
    "    ]\n",
    ")\n",
    "\n",
    "int_pipeline = Pipeline(\n",
    "    steps=[\n",
    "        ('int_selector', FeatureSelector(['volume', 'show', 'run', 'pages', 'update', 'year'])),\n",
    "        ('int_transformer', ToIntTransformer())\n",
    "    ]\n",
    ")\n",
    "\n",
    "cat_pipeline = Pipeline(\n",
    "    steps=[\n",
    "        ('cat_selector', FeatureSelector(['cuzov', 'fuel', 'pages', 'region', 'update'])),\n",
    "        ('cat_transformer', CatTransformer()),\n",
    "        ('cat_encoder', OneHotEncoder())\n",
    "    ]\n",
    ")\n",
    "\n",
    "color_pipeline = Pipeline(\n",
    "    steps=[\n",
    "        ('color_selector', FeatureSelector(['color'])),\n",
    "        ('color_translator', Translator(\n",
    "            [ColumnTranslation(\n",
    "                column_name='color',\n",
    "                to_save=['черный', 'серебристый', 'синий', 'серый', 'белый'],\n",
    "                default='другой'\n",
    "            )])),\n",
    "        ('color_encoder', OneHotEncoder())\n",
    "    ]\n",
    ")\n",
    "\n",
    "no_proc_pipeline = Pipeline(\n",
    "    steps=[\n",
    "        ('no_proc_selector', FeatureSelector(['drive-unit', 'state', 'transmission'])),\n",
    "        ('no_proc_imputer', SimpleImputer(strategy='most_frequent')),\n",
    "        ('no_proc_encoder', OneHotEncoder())\n",
    "    ]\n",
    ")\n",
    "\n",
    "pipeline = FeatureUnion(transformer_list= [\n",
    "    ('feat', feat_pipeline),\n",
    "    ('int', int_pipeline),\n",
    "    ('cat', cat_pipeline),\n",
    "    ('color', color_pipeline),\n",
    "    ('no_proc', no_proc_pipeline)\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>color</th>\n",
       "      <th>cuzov</th>\n",
       "      <th>drive-unit</th>\n",
       "      <th>fuel</th>\n",
       "      <th>pages</th>\n",
       "      <th>region</th>\n",
       "      <th>run</th>\n",
       "      <th>show</th>\n",
       "      <th>state</th>\n",
       "      <th>table</th>\n",
       "      <th>transmission</th>\n",
       "      <th>update</th>\n",
       "      <th>volume</th>\n",
       "      <th>year</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>2025</td>\n",
       "      <td>белый</td>\n",
       "      <td>седан</td>\n",
       "      <td>Задний</td>\n",
       "      <td>бензин</td>\n",
       "      <td>BMW 3 серия E90, E91, E92, E93</td>\n",
       "      <td>Гродно, Гродненская обл.</td>\n",
       "      <td>140900 км</td>\n",
       "      <td>69                         просмотров</td>\n",
       "      <td>с пробегом</td>\n",
       "      <td>Дополнительные опции:\\n                       ...</td>\n",
       "      <td>Автомат</td>\n",
       "      <td>Опубликовано\\n                            \\n  ...</td>\n",
       "      <td>2000 см3</td>\n",
       "      <td>2007</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>594</td>\n",
       "      <td>синий</td>\n",
       "      <td>универсал</td>\n",
       "      <td>Передний</td>\n",
       "      <td>дизель</td>\n",
       "      <td>Opel Vectra C (рестайлинг)</td>\n",
       "      <td>Минск</td>\n",
       "      <td>369645 км</td>\n",
       "      <td>112                         просмотров</td>\n",
       "      <td>с пробегом</td>\n",
       "      <td>Дополнительные опции:\\n                       ...</td>\n",
       "      <td>Механика</td>\n",
       "      <td>Опубликовано\\n                            \\n  ...</td>\n",
       "      <td>1900 см3</td>\n",
       "      <td>2005</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1558</td>\n",
       "      <td>серебристый</td>\n",
       "      <td>минивэн</td>\n",
       "      <td>Передний</td>\n",
       "      <td>дизель</td>\n",
       "      <td>Peugeot 5008 I</td>\n",
       "      <td>Речица, Гомельская обл.</td>\n",
       "      <td>155000 км</td>\n",
       "      <td>81                         просмотр\\n         ...</td>\n",
       "      <td>с пробегом</td>\n",
       "      <td>Дополнительные опции:\\n                       ...</td>\n",
       "      <td>Механика</td>\n",
       "      <td>Опубликовано\\n                            \\n  ...</td>\n",
       "      <td>1600 см3</td>\n",
       "      <td>2011</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9427</td>\n",
       "      <td>другой</td>\n",
       "      <td>седан</td>\n",
       "      <td>Передний</td>\n",
       "      <td>бензин</td>\n",
       "      <td>Opel Vectra A</td>\n",
       "      <td>Гомель, Гомельская обл.</td>\n",
       "      <td>285299 км</td>\n",
       "      <td>73                         просмотра</td>\n",
       "      <td>с пробегом</td>\n",
       "      <td>Дополнительные опции:\\n                       ...</td>\n",
       "      <td>Механика</td>\n",
       "      <td>Опубликовано\\n                            \\n  ...</td>\n",
       "      <td>1800 см3</td>\n",
       "      <td>1991</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10120</td>\n",
       "      <td>черный</td>\n",
       "      <td>седан</td>\n",
       "      <td>Передний</td>\n",
       "      <td>бензин</td>\n",
       "      <td>Mazda 3 BK</td>\n",
       "      <td>Минск</td>\n",
       "      <td>186132 км</td>\n",
       "      <td>420                         просмотров</td>\n",
       "      <td>с пробегом</td>\n",
       "      <td>Дополнительные опции:\\n                       ...</td>\n",
       "      <td>Автомат</td>\n",
       "      <td>Опубликовано\\n                            \\n  ...</td>\n",
       "      <td>2000 см3</td>\n",
       "      <td>2005</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5734</td>\n",
       "      <td>белый</td>\n",
       "      <td>внедорожник 5 дв.</td>\n",
       "      <td>Передний</td>\n",
       "      <td>бензин</td>\n",
       "      <td>Hyundai Tucson III Sport</td>\n",
       "      <td>Минск</td>\n",
       "      <td>44600 км</td>\n",
       "      <td>366                         просмотров\\n      ...</td>\n",
       "      <td>с пробегом</td>\n",
       "      <td>Дополнительные опции:\\n                       ...</td>\n",
       "      <td>Автомат</td>\n",
       "      <td>Опубликовано\\n                            \\n  ...</td>\n",
       "      <td>1600 см3</td>\n",
       "      <td>2017</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5191</td>\n",
       "      <td>синий</td>\n",
       "      <td>седан</td>\n",
       "      <td>Передний</td>\n",
       "      <td>бензин</td>\n",
       "      <td>Opel Vectra B (рестайлинг)</td>\n",
       "      <td>Гомель, Гомельская обл.</td>\n",
       "      <td>346889 км</td>\n",
       "      <td>146                         просмотров\\n      ...</td>\n",
       "      <td>с пробегом</td>\n",
       "      <td>Дополнительные опции:\\n                       ...</td>\n",
       "      <td>Механика</td>\n",
       "      <td>Опубликовано\\n                            \\n  ...</td>\n",
       "      <td>1600 см3</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5390</td>\n",
       "      <td>белый</td>\n",
       "      <td>седан</td>\n",
       "      <td>Передний</td>\n",
       "      <td>дизель</td>\n",
       "      <td>Toyota Avensis I (T220)</td>\n",
       "      <td>Брест, Брестская обл.</td>\n",
       "      <td>391000 км</td>\n",
       "      <td>37                         просмотров</td>\n",
       "      <td>с пробегом</td>\n",
       "      <td>Дополнительные опции:\\n                       ...</td>\n",
       "      <td>Механика</td>\n",
       "      <td>Опубликовано\\n                            \\n  ...</td>\n",
       "      <td>2000 см3</td>\n",
       "      <td>1999</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>860</td>\n",
       "      <td>черный</td>\n",
       "      <td>седан</td>\n",
       "      <td>Передний</td>\n",
       "      <td>бензин</td>\n",
       "      <td>Volvo S80 I Т6</td>\n",
       "      <td>Каменец, Брестская обл.</td>\n",
       "      <td>440000 км</td>\n",
       "      <td>154                         просмотра\\n       ...</td>\n",
       "      <td>с пробегом</td>\n",
       "      <td>Дополнительные опции:\\n                       ...</td>\n",
       "      <td>Автомат</td>\n",
       "      <td>Обновлено\\n                            \\n     ...</td>\n",
       "      <td>2800 см3</td>\n",
       "      <td>1999</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7270</td>\n",
       "      <td>серебристый</td>\n",
       "      <td>хэтчбек 5 дв.</td>\n",
       "      <td>Передний</td>\n",
       "      <td>бензин</td>\n",
       "      <td>Ford Fiesta VI</td>\n",
       "      <td>Минск</td>\n",
       "      <td>225708 км</td>\n",
       "      <td>18                         просмотров</td>\n",
       "      <td>с пробегом</td>\n",
       "      <td>Дополнительные опции:\\n                       ...</td>\n",
       "      <td>Механика</td>\n",
       "      <td>Опубликовано\\n                            \\n  ...</td>\n",
       "      <td>1200 см3</td>\n",
       "      <td>2010</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>8810 rows × 14 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "             color              cuzov drive-unit    fuel  \\\n",
       "2025         белый              седан     Задний  бензин   \n",
       "594          синий          универсал   Передний  дизель   \n",
       "1558   серебристый            минивэн   Передний  дизель   \n",
       "9427        другой              седан   Передний  бензин   \n",
       "10120       черный              седан   Передний  бензин   \n",
       "...            ...                ...        ...     ...   \n",
       "5734         белый  внедорожник 5 дв.   Передний  бензин   \n",
       "5191         синий              седан   Передний  бензин   \n",
       "5390         белый              седан   Передний  дизель   \n",
       "860         черный              седан   Передний  бензин   \n",
       "7270   серебристый      хэтчбек 5 дв.   Передний  бензин   \n",
       "\n",
       "                                pages                    region        run  \\\n",
       "2025   BMW 3 серия E90, E91, E92, E93  Гродно, Гродненская обл.  140900 км   \n",
       "594        Opel Vectra C (рестайлинг)                     Минск  369645 км   \n",
       "1558                   Peugeot 5008 I   Речица, Гомельская обл.  155000 км   \n",
       "9427                    Opel Vectra A   Гомель, Гомельская обл.  285299 км   \n",
       "10120                      Mazda 3 BK                     Минск  186132 км   \n",
       "...                               ...                       ...        ...   \n",
       "5734         Hyundai Tucson III Sport                     Минск   44600 км   \n",
       "5191       Opel Vectra B (рестайлинг)   Гомель, Гомельская обл.  346889 км   \n",
       "5390          Toyota Avensis I (T220)     Брест, Брестская обл.  391000 км   \n",
       "860                    Volvo S80 I Т6   Каменец, Брестская обл.  440000 км   \n",
       "7270                   Ford Fiesta VI                     Минск  225708 км   \n",
       "\n",
       "                                                    show       state  \\\n",
       "2025               69                         просмотров  с пробегом   \n",
       "594               112                         просмотров  с пробегом   \n",
       "1558   81                         просмотр\\n         ...  с пробегом   \n",
       "9427                73                         просмотра  с пробегом   \n",
       "10120             420                         просмотров  с пробегом   \n",
       "...                                                  ...         ...   \n",
       "5734   366                         просмотров\\n      ...  с пробегом   \n",
       "5191   146                         просмотров\\n      ...  с пробегом   \n",
       "5390               37                         просмотров  с пробегом   \n",
       "860    154                         просмотра\\n       ...  с пробегом   \n",
       "7270               18                         просмотров  с пробегом   \n",
       "\n",
       "                                                   table transmission  \\\n",
       "2025   Дополнительные опции:\\n                       ...      Автомат   \n",
       "594    Дополнительные опции:\\n                       ...     Механика   \n",
       "1558   Дополнительные опции:\\n                       ...     Механика   \n",
       "9427   Дополнительные опции:\\n                       ...     Механика   \n",
       "10120  Дополнительные опции:\\n                       ...      Автомат   \n",
       "...                                                  ...          ...   \n",
       "5734   Дополнительные опции:\\n                       ...      Автомат   \n",
       "5191   Дополнительные опции:\\n                       ...     Механика   \n",
       "5390   Дополнительные опции:\\n                       ...     Механика   \n",
       "860    Дополнительные опции:\\n                       ...      Автомат   \n",
       "7270   Дополнительные опции:\\n                       ...     Механика   \n",
       "\n",
       "                                                  update    volume  year  \n",
       "2025   Опубликовано\\n                            \\n  ...  2000 см3  2007  \n",
       "594    Опубликовано\\n                            \\n  ...  1900 см3  2005  \n",
       "1558   Опубликовано\\n                            \\n  ...  1600 см3  2011  \n",
       "9427   Опубликовано\\n                            \\n  ...  1800 см3  1991  \n",
       "10120  Опубликовано\\n                            \\n  ...  2000 см3  2005  \n",
       "...                                                  ...       ...   ...  \n",
       "5734   Опубликовано\\n                            \\n  ...  1600 см3  2017  \n",
       "5191   Опубликовано\\n                            \\n  ...  1600 см3  2000  \n",
       "5390   Опубликовано\\n                            \\n  ...  2000 см3  1999  \n",
       "860    Обновлено\\n                            \\n     ...  2800 см3  1999  \n",
       "7270   Опубликовано\\n                            \\n  ...  1200 см3  2010  \n",
       "\n",
       "[8810 rows x 14 columns]"
      ]
     },
     "execution_count": 124,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.tree import DecisionTreeRegressor\n",
    "from sklearn.ensemble import RandomForestRegressor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Pipeline(memory=None,\n",
       "         steps=[('preprocessing',\n",
       "                 FeatureUnion(n_jobs=None,\n",
       "                              transformer_list=[('feat',\n",
       "                                                 Pipeline(memory=None,\n",
       "                                                          steps=[('feat_selector',\n",
       "                                                                  FeatureSelector(feature_names=['table'])),\n",
       "                                                                 ('feat_generator',\n",
       "                                                                  FeatureGenerator())],\n",
       "                                                          verbose=False)),\n",
       "                                                ('int',\n",
       "                                                 Pipeline(memory=None,\n",
       "                                                          steps=[('int_selector',\n",
       "                                                                  FeatureSelector(feature_names=['volume',\n",
       "                                                                                                 'show',\n",
       "                                                                                                 'run',\n",
       "                                                                                                 'pages',...\n",
       "                              transformer_weights=None, verbose=False)),\n",
       "                ('model',\n",
       "                 RandomForestRegressor(bootstrap=True, criterion='mse',\n",
       "                                       max_depth=20, max_features=None,\n",
       "                                       max_leaf_nodes=None,\n",
       "                                       min_impurity_decrease=0.0,\n",
       "                                       min_impurity_split=None,\n",
       "                                       min_samples_leaf=1, min_samples_split=5,\n",
       "                                       min_weight_fraction_leaf=0.0,\n",
       "                                       n_estimators=150, n_jobs=-1,\n",
       "                                       oob_score=False, random_state=42,\n",
       "                                       verbose=0, warm_start=False))],\n",
       "         verbose=False)"
      ]
     },
     "execution_count": 131,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pip_rf = Pipeline(\n",
    "    steps=[\n",
    "        ('preprocessing', pipeline),\n",
    "        ('model', RandomForestRegressor(\n",
    "                        max_depth=20, \n",
    "                        min_samples_split=5, \n",
    "                        min_samples_leaf=1,\n",
    "                        max_features=None,\n",
    "                        random_state=42,\n",
    "                        n_estimators=150,\n",
    "                        n_jobs=-1\n",
    "        ))\n",
    "    ]\n",
    ")\n",
    "\n",
    "pip_rf.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8382482476676113\n"
     ]
    }
   ],
   "source": [
    "y_pred = pip_rf.predict(X_test)\n",
    "print(metrics.r2_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([2.55789054e-05, 2.28193229e-03, 2.42926457e-03, 5.25828851e-03,\n",
       "       2.41640553e-03, 1.31000673e-03, 0.00000000e+00, 1.04985789e-04,\n",
       "       1.34394164e-04, 5.33885113e-03, 3.11355395e-04, 1.54794553e-04,\n",
       "       1.35649684e-03, 0.00000000e+00, 2.57006618e-05, 4.67825080e-05,\n",
       "       4.97283369e-06, 2.96438511e-04, 3.85755066e-05, 7.97466164e-05,\n",
       "       1.20068329e-05, 6.98040176e-05, 4.72782961e-06, 0.00000000e+00,\n",
       "       1.69452888e-04, 1.78948246e-03, 1.29088471e-04, 7.26648525e-05,\n",
       "       1.58938534e-04, 3.70763685e-04, 0.00000000e+00, 1.74991629e-04,\n",
       "       1.17369843e-04, 5.42174881e-04, 4.57444204e-04, 4.16089870e-05,\n",
       "       1.00851820e-04, 2.19110825e-03, 0.00000000e+00, 4.94666468e-05,\n",
       "       3.10768022e-01, 2.24410613e-02, 1.94468095e-02, 2.41365197e-03,\n",
       "       1.26694827e-03, 3.32575387e-04, 2.46083393e-05, 7.37352580e-05,\n",
       "       5.71481172e-01, 2.95096635e-03, 0.00000000e+00, 2.93524769e-04,\n",
       "       1.25859849e-03, 0.00000000e+00, 0.00000000e+00, 4.52431045e-05,\n",
       "       5.29787799e-03, 1.75338482e-04, 0.00000000e+00, 0.00000000e+00,\n",
       "       5.41247918e-04, 5.12650674e-06, 4.28492000e-04, 5.51896797e-03,\n",
       "       1.50984936e-03, 3.90907579e-04, 9.79301231e-05, 0.00000000e+00,\n",
       "       0.00000000e+00, 6.68336664e-06, 0.00000000e+00, 0.00000000e+00,\n",
       "       3.82462494e-04, 4.61586147e-04, 6.01868296e-05, 3.19960928e-03,\n",
       "       0.00000000e+00, 2.36358483e-05, 1.07037444e-04, 0.00000000e+00,\n",
       "       3.92225091e-04, 2.08918888e-05, 8.04426346e-04, 3.22264828e-04,\n",
       "       3.22200435e-03, 4.21953301e-04, 4.21262482e-06, 1.72701104e-05,\n",
       "       5.99216437e-06, 0.00000000e+00, 1.56805378e-05, 2.95899538e-05,\n",
       "       3.28792535e-05, 2.39924233e-03, 1.65727098e-03, 0.00000000e+00,\n",
       "       4.32637421e-04, 0.00000000e+00, 4.80557067e-04, 8.70096871e-04,\n",
       "       1.57712529e-03, 4.76022457e-05, 5.87924406e-03, 0.00000000e+00,\n",
       "       2.13879987e-04, 5.44110263e-04, 0.00000000e+00, 1.53846976e-03])"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pip_tree['model'].feature_importances_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import GridSearchCV"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 119,
   "metadata": {},
   "outputs": [],
   "source": [
    "param_grid = [\n",
    "    {\n",
    "        'max_depth': [ 8, 10, 15],\n",
    "        'n_estimators': [100, 150],\n",
    "        'min_samples_split': [3, 5]\n",
    "    }\n",
    "]\n",
    "\n",
    "pip_rf = Pipeline(\n",
    "    steps=[\n",
    "        ('preprocessing', pipeline),\n",
    "        ('rf', GridSearchCV(RandomForestRegressor(\n",
    "                        random_state=42,\n",
    "                        n_jobs=-1\n",
    "        ), param_grid, scoring='r2', cv=5))\n",
    "    ]\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8419051152075978\n"
     ]
    }
   ],
   "source": [
    "pip_rf.fit(X_train, y_train)\n",
    "y_true, y_pred = y_test, pip_rf.predict(X_test)\n",
    "print(metrics.r2_score(y_true, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomForestRegressor(bootstrap=True, criterion='mse', max_depth=15,\n",
       "                      max_features='auto', max_leaf_nodes=None,\n",
       "                      min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "                      min_samples_leaf=1, min_samples_split=3,\n",
       "                      min_weight_fraction_leaf=0.0, n_estimators=100, n_jobs=-1,\n",
       "                      oob_score=False, random_state=42, verbose=0,\n",
       "                      warm_start=False)"
      ]
     },
     "execution_count": 130,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pip_rf[1].best_estimator_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "metadata": {},
   "outputs": [],
   "source": [
    "param_grid = {\"criterion\": [\"mse\", \"mae\"],\n",
    "              \"min_samples_split\": [5, 10, 20],\n",
    "              \"max_depth\": [2, 5, 8, 10],\n",
    "               \"min_samples_leaf\": [20, 40, 100],\n",
    "               \"max_leaf_nodes\": [None, 5, 20, 10],\n",
    "              }\n",
    "pip_tree = Pipeline(\n",
    "    steps=[\n",
    "        ('preprocessing', pipeline),\n",
    "        ('tree', GridSearchCV(DecisionTreeRegressor(\n",
    "                        random_state=42\n",
    "        ), param_grid, scoring='r2', cv=5))\n",
    "    ]\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8382482476676113\n"
     ]
    }
   ],
   "source": [
    "pip_tree.fit(X_train, y_train)\n",
    "y_true, y_pred = y_test, pip_rf.predict(X_test)\n",
    "print(metrics.r2_score(y_true, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DecisionTreeRegressor(criterion='mse', max_depth=10, max_features=None,\n",
       "                      max_leaf_nodes=None, min_impurity_decrease=0.0,\n",
       "                      min_impurity_split=None, min_samples_leaf=20,\n",
       "                      min_samples_split=5, min_weight_fraction_leaf=0.0,\n",
       "                      presort=False, random_state=42, splitter='best')"
      ]
     },
     "execution_count": 141,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pip_tree[1].best_estimator_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}