baseline_boosters.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from math import sin, cos, sqrt, atan2, radians\n",
    "from tqdm import tqdm\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.metrics import mean_squared_error\n",
    "import lightgbm as lgb\n",
    "from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def rmse(y_true, y_pred):\n",
    "    return sqrt(mean_squared_error(y_true, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "train = pd.read_csv('data/train.csv', index_col=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "test = pd.read_csv('data/test.csv', index_col=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "train['isTrain'] = True\n",
    "test['isTrain'] = False"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = train.append(test, sort=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "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>count</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>min</th>\n",
       "      <th>25%</th>\n",
       "      <th>50%</th>\n",
       "      <th>75%</th>\n",
       "      <th>max</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>id</th>\n",
       "      <td>8765.0</td>\n",
       "      <td>4404.586366</td>\n",
       "      <td>2542.532548</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>2207.000000</td>\n",
       "      <td>4403.000000</td>\n",
       "      <td>6603.000000</td>\n",
       "      <td>8811.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>atm_group</th>\n",
       "      <td>8765.0</td>\n",
       "      <td>4407.008842</td>\n",
       "      <td>2427.934099</td>\n",
       "      <td>32.000000</td>\n",
       "      <td>1942.000000</td>\n",
       "      <td>5478.000000</td>\n",
       "      <td>5478.000000</td>\n",
       "      <td>8083.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>lat</th>\n",
       "      <td>8345.0</td>\n",
       "      <td>54.138304</td>\n",
       "      <td>4.329293</td>\n",
       "      <td>42.057080</td>\n",
       "      <td>52.243997</td>\n",
       "      <td>55.060936</td>\n",
       "      <td>56.116673</td>\n",
       "      <td>69.495093</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>long</th>\n",
       "      <td>8345.0</td>\n",
       "      <td>63.500873</td>\n",
       "      <td>32.787366</td>\n",
       "      <td>19.941530</td>\n",
       "      <td>37.851107</td>\n",
       "      <td>49.071694</td>\n",
       "      <td>84.950218</td>\n",
       "      <td>173.706490</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>target</th>\n",
       "      <td>6261.0</td>\n",
       "      <td>0.000715</td>\n",
       "      <td>0.085852</td>\n",
       "      <td>-0.145001</td>\n",
       "      <td>-0.061176</td>\n",
       "      <td>-0.015573</td>\n",
       "      <td>0.040149</td>\n",
       "      <td>0.218608</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            count         mean          std        min          25%  \\\n",
       "id         8765.0  4404.586366  2542.532548   1.000000  2207.000000   \n",
       "atm_group  8765.0  4407.008842  2427.934099  32.000000  1942.000000   \n",
       "lat        8345.0    54.138304     4.329293  42.057080    52.243997   \n",
       "long       8345.0    63.500873    32.787366  19.941530    37.851107   \n",
       "target     6261.0     0.000715     0.085852  -0.145001    -0.061176   \n",
       "\n",
       "                   50%          75%          max  \n",
       "id         4403.000000  6603.000000  8811.000000  \n",
       "atm_group  5478.000000  5478.000000  8083.000000  \n",
       "lat          55.060936    56.116673    69.495093  \n",
       "long         49.071694    84.950218   173.706490  \n",
       "target       -0.015573     0.040149     0.218608  "
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X.describe().T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "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>count</th>\n",
       "      <th>unique</th>\n",
       "      <th>top</th>\n",
       "      <th>freq</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>address</th>\n",
       "      <td>8765</td>\n",
       "      <td>7481</td>\n",
       "      <td>HOROSHEVSKOE SH,V38       MOSKVA</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>address_rus</th>\n",
       "      <td>8345</td>\n",
       "      <td>6147</td>\n",
       "      <td>квартал В-13, городской округ Волгодонск, Рост...</td>\n",
       "      <td>10</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            count unique                                                top  \\\n",
       "address      8765   7481             HOROSHEVSKOE SH,V38       MOSKVA         \n",
       "address_rus  8345   6147  квартал В-13, городской округ Волгодонск, Рост...   \n",
       "\n",
       "            freq  \n",
       "address        7  \n",
       "address_rus   10  "
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X.describe(include='object').T"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Количество банкоматов стоящих по одному адресу"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "counts = X.groupby('address_rus')['id'].count().reset_index().rename(columns={'id':'count'})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = pd.merge(X, counts, how='left', on='address_rus')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "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>id</th>\n",
       "      <th>atm_group</th>\n",
       "      <th>address</th>\n",
       "      <th>address_rus</th>\n",
       "      <th>lat</th>\n",
       "      <th>long</th>\n",
       "      <th>target</th>\n",
       "      <th>isTrain</th>\n",
       "      <th>count</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>8526.0</td>\n",
       "      <td>32.0</td>\n",
       "      <td>EMELYANOVA,34          Y-SAKHALINSK</td>\n",
       "      <td>улица А.О. Емельянова, 34, Южно-Сахалинск, Сах...</td>\n",
       "      <td>46.940995</td>\n",
       "      <td>142.738319</td>\n",
       "      <td>0.011500</td>\n",
       "      <td>True</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>8532.0</td>\n",
       "      <td>32.0</td>\n",
       "      <td>KOMSOMOLSKAYA,259B     Y.SAKHALINSK</td>\n",
       "      <td>Комсомольская улица, 259, Южно-Сахалинск, Саха...</td>\n",
       "      <td>46.937353</td>\n",
       "      <td>142.753348</td>\n",
       "      <td>0.029710</td>\n",
       "      <td>True</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>8533.0</td>\n",
       "      <td>32.0</td>\n",
       "      <td>KOMMUN. PR., 32        YUZHNO SAKHAL</td>\n",
       "      <td>Коммунистический проспект, Южно-Сахалинск, Сах...</td>\n",
       "      <td>46.959413</td>\n",
       "      <td>142.741113</td>\n",
       "      <td>0.009540</td>\n",
       "      <td>True</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>8684.0</td>\n",
       "      <td>32.0</td>\n",
       "      <td>LENINGRADSKIY PR.,76A  MOSCOW</td>\n",
       "      <td>Ленинградский проспект, 76А, Москва, Россия, 1...</td>\n",
       "      <td>55.805827</td>\n",
       "      <td>37.515146</td>\n",
       "      <td>-0.094035</td>\n",
       "      <td>True</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>37.0</td>\n",
       "      <td>32.0</td>\n",
       "      <td>GVARDEYSKAYA PL., 2    NORILSK</td>\n",
       "      <td>Гвардейская площадь, 2, Норильск, Красноярский...</td>\n",
       "      <td>69.343541</td>\n",
       "      <td>88.211228</td>\n",
       "      <td>0.079277</td>\n",
       "      <td>True</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       id  atm_group                                 address  \\\n",
       "0  8526.0       32.0  EMELYANOVA,34          Y-SAKHALINSK      \n",
       "1  8532.0       32.0  KOMSOMOLSKAYA,259B     Y.SAKHALINSK      \n",
       "2  8533.0       32.0  KOMMUN. PR., 32        YUZHNO SAKHAL     \n",
       "3  8684.0       32.0  LENINGRADSKIY PR.,76A  MOSCOW            \n",
       "4    37.0       32.0  GVARDEYSKAYA PL., 2    NORILSK           \n",
       "\n",
       "                                         address_rus        lat        long  \\\n",
       "0  улица А.О. Емельянова, 34, Южно-Сахалинск, Сах...  46.940995  142.738319   \n",
       "1  Комсомольская улица, 259, Южно-Сахалинск, Саха...  46.937353  142.753348   \n",
       "2  Коммунистический проспект, Южно-Сахалинск, Сах...  46.959413  142.741113   \n",
       "3  Ленинградский проспект, 76А, Москва, Россия, 1...  55.805827   37.515146   \n",
       "4  Гвардейская площадь, 2, Норильск, Красноярский...  69.343541   88.211228   \n",
       "\n",
       "     target  isTrain  count  \n",
       "0  0.011500     True    1.0  \n",
       "1  0.029710     True    1.0  \n",
       "2  0.009540     True    1.0  \n",
       "3 -0.094035     True    1.0  \n",
       "4  0.079277     True    1.0  "
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Расстояния до ближайших банкоматов"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "R = 6373.0 # радиус земли в километрах\n",
    "\n",
    "def distance(x,y):\n",
    "    \"\"\"\n",
    "    Параметры\n",
    "    ----------\n",
    "    x : tuple, широта и долгота первой геокоординаты \n",
    "    y : tuple, широта и долгота второй геокоординаты \n",
    "    \n",
    "    Результат\n",
    "    ----------\n",
    "    result : дистанция в километрах между двумя геокоординатами\n",
    "    \"\"\"\n",
    "    lat_a, long_a, lat_b, long_b = map(radians, [*x,*y])    \n",
    "    dlon = long_b - long_a\n",
    "    dlat = lat_b - lat_a\n",
    "    a = sin(dlat/2)**2 + cos(lat_a) * cos(lat_b) * sin(dlon/2)**2\n",
    "    c = 2 * atan2(sqrt(a), sqrt(1 - a))\n",
    "    return R * c"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "knc = KNeighborsClassifier(metric=distance)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "dots = X[['lat','long']].dropna()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "KNeighborsClassifier(algorithm='auto', leaf_size=30,\n",
       "           metric=<function distance at 0x7f3f7cb0f9d8>,\n",
       "           metric_params=None, n_jobs=None, n_neighbors=5, p=2,\n",
       "           weights='uniform')"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "knc.fit(X=dots , y=np.ones(dots.shape[0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "distances, indexes = knc.kneighbors(X=dots,n_neighbors=6,)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "for i in range(1,6):\n",
    "    dots['distance_%s'%i] = distances[:,i]\n",
    "    dots['indexes_%s'%i] = indexes[:,i]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "dots['mean'] = dots.iloc[:,dots.columns.str.contains('distance')].mean(axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = pd.concat([X,dots], axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Город размещения"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "X['city'] = X[~X.address_rus.isnull()].address_rus.apply(lambda x: x.split(',')[2]) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "rare_cities = X.city.value_counts()[(X.city.value_counts() < 20) ==True].index"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "X.city = X.city.apply(lambda x: 'RARE' if x in rare_cities else x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "X.city= X.city.rank().fillna(-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_ = X[X.isTrain][['atm_group', 'lat', 'long',  'count', 'distance_1',\n",
    "                    'distance_2',  'distance_3',  'distance_4', 'distance_5',\n",
    "                    'indexes_5', 'mean', 'city']]\n",
    "Y_ = X.loc[X.isTrain, 'target']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_valid, Y_train, Y_valid = train_test_split(X_, Y_, test_size=0.25, random_state=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "gbm = lgb.LGBMRegressor(objective = 'regression',  \n",
    "                            max_depth = 3,\n",
    "                            colsample_bytre = 0.8,\n",
    "                            subsample = 0.8, \n",
    "                            learning_rate = 0.1,\n",
    "                            n_estimators = 300, random_state=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1]\tvalid_0's l2: 0.00627427\tvalid_0's rmse: 0.0792103\n",
      "Training until validation scores don't improve for 5 rounds.\n",
      "[2]\tvalid_0's l2: 0.00555286\tvalid_0's rmse: 0.0745175\n",
      "[3]\tvalid_0's l2: 0.00497472\tvalid_0's rmse: 0.0705317\n",
      "[4]\tvalid_0's l2: 0.00450262\tvalid_0's rmse: 0.0671016\n",
      "[5]\tvalid_0's l2: 0.00411768\tvalid_0's rmse: 0.0641692\n",
      "[6]\tvalid_0's l2: 0.00380687\tvalid_0's rmse: 0.0616998\n",
      "[7]\tvalid_0's l2: 0.003551\tvalid_0's rmse: 0.0595902\n",
      "[8]\tvalid_0's l2: 0.00331214\tvalid_0's rmse: 0.0575512\n",
      "[9]\tvalid_0's l2: 0.00314014\tvalid_0's rmse: 0.056037\n",
      "[10]\tvalid_0's l2: 0.00300172\tvalid_0's rmse: 0.054788\n",
      "[11]\tvalid_0's l2: 0.00288327\tvalid_0's rmse: 0.0536961\n",
      "[12]\tvalid_0's l2: 0.00276677\tvalid_0's rmse: 0.0526001\n",
      "[13]\tvalid_0's l2: 0.00268019\tvalid_0's rmse: 0.0517706\n",
      "[14]\tvalid_0's l2: 0.00259976\tvalid_0's rmse: 0.0509879\n",
      "[15]\tvalid_0's l2: 0.0025416\tvalid_0's rmse: 0.0504143\n",
      "[16]\tvalid_0's l2: 0.00249533\tvalid_0's rmse: 0.0499533\n",
      "[17]\tvalid_0's l2: 0.00245775\tvalid_0's rmse: 0.0495757\n",
      "[18]\tvalid_0's l2: 0.00241735\tvalid_0's rmse: 0.0491666\n",
      "[19]\tvalid_0's l2: 0.00238695\tvalid_0's rmse: 0.0488564\n",
      "[20]\tvalid_0's l2: 0.00235907\tvalid_0's rmse: 0.0485702\n",
      "[21]\tvalid_0's l2: 0.00234164\tvalid_0's rmse: 0.0483905\n",
      "[22]\tvalid_0's l2: 0.00232287\tvalid_0's rmse: 0.0481961\n",
      "[23]\tvalid_0's l2: 0.0023021\tvalid_0's rmse: 0.0479803\n",
      "[24]\tvalid_0's l2: 0.00229276\tvalid_0's rmse: 0.0478828\n",
      "[25]\tvalid_0's l2: 0.00228011\tvalid_0's rmse: 0.0477505\n",
      "[26]\tvalid_0's l2: 0.00226566\tvalid_0's rmse: 0.047599\n",
      "[27]\tvalid_0's l2: 0.00225853\tvalid_0's rmse: 0.047524\n",
      "[28]\tvalid_0's l2: 0.00225087\tvalid_0's rmse: 0.0474433\n",
      "[29]\tvalid_0's l2: 0.00224508\tvalid_0's rmse: 0.0473823\n",
      "[30]\tvalid_0's l2: 0.00223679\tvalid_0's rmse: 0.0472947\n",
      "[31]\tvalid_0's l2: 0.00222833\tvalid_0's rmse: 0.0472051\n",
      "[32]\tvalid_0's l2: 0.00222035\tvalid_0's rmse: 0.0471206\n",
      "[33]\tvalid_0's l2: 0.00221483\tvalid_0's rmse: 0.047062\n",
      "[34]\tvalid_0's l2: 0.00220896\tvalid_0's rmse: 0.0469996\n",
      "[35]\tvalid_0's l2: 0.00220691\tvalid_0's rmse: 0.0469778\n",
      "[36]\tvalid_0's l2: 0.00220432\tvalid_0's rmse: 0.0469502\n",
      "[37]\tvalid_0's l2: 0.00220106\tvalid_0's rmse: 0.0469154\n",
      "[38]\tvalid_0's l2: 0.00219872\tvalid_0's rmse: 0.0468905\n",
      "[39]\tvalid_0's l2: 0.00219181\tvalid_0's rmse: 0.0468167\n",
      "[40]\tvalid_0's l2: 0.002189\tvalid_0's rmse: 0.0467867\n",
      "[41]\tvalid_0's l2: 0.00218744\tvalid_0's rmse: 0.0467701\n",
      "[42]\tvalid_0's l2: 0.00218344\tvalid_0's rmse: 0.0467273\n",
      "[43]\tvalid_0's l2: 0.00218383\tvalid_0's rmse: 0.0467315\n",
      "[44]\tvalid_0's l2: 0.00218073\tvalid_0's rmse: 0.0466983\n",
      "[45]\tvalid_0's l2: 0.00217606\tvalid_0's rmse: 0.0466482\n",
      "[46]\tvalid_0's l2: 0.00216955\tvalid_0's rmse: 0.0465784\n",
      "[47]\tvalid_0's l2: 0.00216863\tvalid_0's rmse: 0.0465685\n",
      "[48]\tvalid_0's l2: 0.00216715\tvalid_0's rmse: 0.0465527\n",
      "[49]\tvalid_0's l2: 0.00216719\tvalid_0's rmse: 0.0465531\n",
      "[50]\tvalid_0's l2: 0.00216544\tvalid_0's rmse: 0.0465343\n",
      "[51]\tvalid_0's l2: 0.00216192\tvalid_0's rmse: 0.0464965\n",
      "[52]\tvalid_0's l2: 0.00215481\tvalid_0's rmse: 0.0464199\n",
      "[53]\tvalid_0's l2: 0.00214826\tvalid_0's rmse: 0.0463494\n",
      "[54]\tvalid_0's l2: 0.00214613\tvalid_0's rmse: 0.0463263\n",
      "[55]\tvalid_0's l2: 0.00214204\tvalid_0's rmse: 0.0462822\n",
      "[56]\tvalid_0's l2: 0.00214296\tvalid_0's rmse: 0.0462922\n",
      "[57]\tvalid_0's l2: 0.00214103\tvalid_0's rmse: 0.0462713\n",
      "[58]\tvalid_0's l2: 0.00213991\tvalid_0's rmse: 0.0462592\n",
      "[59]\tvalid_0's l2: 0.00213757\tvalid_0's rmse: 0.0462338\n",
      "[60]\tvalid_0's l2: 0.00213714\tvalid_0's rmse: 0.0462292\n",
      "[61]\tvalid_0's l2: 0.00213778\tvalid_0's rmse: 0.0462361\n",
      "[62]\tvalid_0's l2: 0.00213309\tvalid_0's rmse: 0.0461854\n",
      "[63]\tvalid_0's l2: 0.00213313\tvalid_0's rmse: 0.0461859\n",
      "[64]\tvalid_0's l2: 0.00213091\tvalid_0's rmse: 0.0461618\n",
      "[65]\tvalid_0's l2: 0.00212864\tvalid_0's rmse: 0.0461372\n",
      "[66]\tvalid_0's l2: 0.002126\tvalid_0's rmse: 0.0461086\n",
      "[67]\tvalid_0's l2: 0.00212662\tvalid_0's rmse: 0.0461153\n",
      "[68]\tvalid_0's l2: 0.00212422\tvalid_0's rmse: 0.0460892\n",
      "[69]\tvalid_0's l2: 0.00212268\tvalid_0's rmse: 0.0460726\n",
      "[70]\tvalid_0's l2: 0.00211657\tvalid_0's rmse: 0.0460062\n",
      "[71]\tvalid_0's l2: 0.00211623\tvalid_0's rmse: 0.0460025\n",
      "[72]\tvalid_0's l2: 0.00211586\tvalid_0's rmse: 0.0459984\n",
      "[73]\tvalid_0's l2: 0.00211414\tvalid_0's rmse: 0.0459798\n",
      "[74]\tvalid_0's l2: 0.00211235\tvalid_0's rmse: 0.0459603\n",
      "[75]\tvalid_0's l2: 0.00211152\tvalid_0's rmse: 0.0459513\n",
      "[76]\tvalid_0's l2: 0.00211071\tvalid_0's rmse: 0.0459425\n",
      "[77]\tvalid_0's l2: 0.00210726\tvalid_0's rmse: 0.0459049\n",
      "[78]\tvalid_0's l2: 0.00210763\tvalid_0's rmse: 0.045909\n",
      "[79]\tvalid_0's l2: 0.00210864\tvalid_0's rmse: 0.0459199\n",
      "[80]\tvalid_0's l2: 0.00210884\tvalid_0's rmse: 0.0459221\n",
      "[81]\tvalid_0's l2: 0.00210912\tvalid_0's rmse: 0.0459251\n",
      "[82]\tvalid_0's l2: 0.00210736\tvalid_0's rmse: 0.045906\n",
      "Early stopping, best iteration is:\n",
      "[77]\tvalid_0's l2: 0.00210726\tvalid_0's rmse: 0.0459049\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "LGBMRegressor(boosting_type='gbdt', class_weight=None, colsample_bytre=0.8,\n",
       "       colsample_bytree=1.0, importance_type='split', learning_rate=0.1,\n",
       "       max_depth=3, min_child_samples=20, min_child_weight=0.001,\n",
       "       min_split_gain=0.0, n_estimators=300, n_jobs=-1, num_leaves=31,\n",
       "       objective='regression', random_state=1, reg_alpha=0.0,\n",
       "       reg_lambda=0.0, silent=True, subsample=0.8,\n",
       "       subsample_for_bin=200000, subsample_freq=0)"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gbm.fit(X_train, Y_train, eval_set=[(X_valid, Y_valid)], eval_metric='rmse', early_stopping_rounds=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7fd4705da358>"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "lgb.plot_importance(gbm)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.045904893145415314"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rmse(Y_valid, gbm.predict(X_valid))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.08463725683992336"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rmse(Y_valid, np.zeros(Y_valid.shape[0]))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}