trainning.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import xgboost as xgb\n",
    "from sklearn.model_selection import train_test_split, RandomizedSearchCV\n",
    "from scipy import stats\n",
    "import seaborn as sns\n",
    "from sklearn.metrics import mean_squared_error\n",
    "from sklearn.ensemble import RandomForestRegressor\n",
    "from xgboost.sklearn import XGBRegressor\n",
    "from sklearn.datasets import make_regression\n",
    "import sklearn\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "np.random.seed(10)\n",
    "import lightgbm as lgb\n",
    "from lightgbm import plot_importance\n",
    "from sklearn.metrics import mean_squared_error\n",
    "from scipy.stats import randint as sp_randint\n",
    "from scipy.stats import randint as sp_uniform"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Carga de los df de features para entrenar y predecir"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "windows = [\"18_20\",\"19_21\",\"20_22\",\"21_23\"]\n",
    "features = {}\n",
    "label_auc = {}\n",
    "label_inst = {}\n",
    "label_clas_auc = {}\n",
    "label_clas_inst = {}\n",
    "\n",
    "for window in windows:\n",
    "    features[window] = pd.read_csv(\"windows/{}/features.csv\".format(window), index_col=\"ref_hash\")\n",
    "    label_auc[window] = pd.read_csv(\"windows/{}/labels_auc.csv\".format(window), index_col=\"ref_hash\")\n",
    "    label_inst[window] = pd.read_csv(\"windows/{}/labels_inst.csv\".format(window), index_col=\"ref_hash\")\n",
    "    label_clas_auc[window] = pd.DataFrame({'ref_hash': label_auc[window].index, 'label_auc': (label_auc[window]['label_auc'] == 259200).astype(int)}).set_index(\"ref_hash\")\n",
    "    label_clas_inst[window] = pd.DataFrame({'ref_hash': label_inst[window].index, 'label_inst': (label_inst[window]['label_inst'] == 259200).astype(int)}).set_index(\"ref_hash\")\n",
    "    \n",
    "features_to_predict = pd.read_csv(\"windows/24_26/features.csv\", index_col=\"ref_hash\") "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Selección de features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def select_features(df, feature_list):\n",
    "    return df.reindex(columns=feature_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "features_list_class_auc = [\n",
    "    'appearances_in_auctions', \n",
    "    'user_appeared_last_day', \n",
    "    'time_to_reappear',\n",
    "    #'amount_of_clicks', \n",
    "    #'has_installed', \n",
    "    #'user_clicked_last_day',\n",
    "    #'user_installed_last_day', \n",
    "    #'amount_of_installs', \n",
    "    #'cant_max_day',\n",
    "    #'cant_min_day', \n",
    "    #'mean_time_to_click', \n",
    "    #'max_time_click',\n",
    "    #'min_time_click', \n",
    "    'mean_auctions_per_day', \n",
    "    #'mean_events_per_day',\n",
    "    #'mean_clicks_per_day', \n",
    "    'amount_auctions_in_last_hour',\n",
    "    'amount_auctions_in_last_2_hours', \n",
    "    'amount_auctions_in_last_5_hours',\n",
    "    'amount_auctions_in_last_12_hours', \n",
    "    #'amount_auctions_in_last_24_hours',\n",
    "    #'amount_events_in_last_hour', \n",
    "    #'amount_events_in_last_2_hours',\n",
    "    #'amount_auctions_in_first_hour', \n",
    "    #'amount_auctions_in_first_3_hours',\n",
    "    #'amount_auctions_in_first_5_hours', \n",
    "    'amount_auctions_in_first_12_hours',\n",
    "    #'amount_events_in_first_hour', \n",
    "    #'amount_events_in_first_5_hours',\n",
    "    'amount_events_in_first_12_hours', \n",
    "    #'amount_clicks_in_last_2_hours',\n",
    "    #'amount_clicks_in_last_4_hours', \n",
    "    'device_os', \n",
    "    #'std_time_to_click',\n",
    "    'std_time_auctions', \n",
    "    #'max_time_install', \n",
    "    #'min_time_install',\n",
    "    #'mean_time_install', \n",
    "    #'std_time_install', \n",
    "    #'max_time_events',\n",
    "    #'min_time_events', \n",
    "    #'mean_time_events', \n",
    "    'std_time_events',\n",
    "    #'installs_per_events', \n",
    "    #'installs_per_clicks', \n",
    "    #'events_x_app_210',\n",
    "    #'events_x_app_122', \n",
    "    #'events_x_app_65', \n",
    "    #'events_x_app_121',\n",
    "    #'events_x_app_26', \n",
    "    #'most_installed_apps_used', \n",
    "    #'cant_apps_used',\n",
    "    #'cant_events_atributed', \n",
    "    #'has_events_atributed',\n",
    "    #'has_events_ids_with_installs', \n",
    "    #'has_events_ids_without_installs',\n",
    "    'cant_events_0_4', \n",
    "    'cant_events_4_8', \n",
    "    'cant_events_8_12',\n",
    "    'cant_events_12_16', \n",
    "    'cant_events_16_20', \n",
    "    'cant_events_20_24',\n",
    "    'cant_auctions_0_4', \n",
    "    'cant_auctions_4_8', \n",
    "    'cant_auctions_8_12',\n",
    "    'cant_auctions_12_16', \n",
    "    'cant_auctions_16_20', \n",
    "    'cant_auctions_20_24',\n",
    "    #'implicit', \n",
    "    #'latitude', \n",
    "    #'longitude', \n",
    "    #'clicked_in_last_5_minutes',\n",
    "    #'clicked_with_wifi_in_last_3_hours', \n",
    "    'hour_install', \n",
    "    'hour_events',\n",
    "    'hour_clicks', \n",
    "    'hour_auctions',\n",
    "    #'amount_auctions_in_last_half_hour',\n",
    "    #'amount_auctions_in_last_15_minutes',\n",
    "    \"mean_appearances_events\",\n",
    "    \"std_appearances_events\",\n",
    "    \"std_appearances_auctions\",\n",
    "    'cant_devices',\n",
    "    \"mean_appearances_auctions\",\n",
    "    \"mean_of_first_auction\",\n",
    "    \"mean_of_first_event\",\n",
    "    \"time_to_first_appearance_in_auctions\",\n",
    "    \"mean_auctions_last_24hs\",\n",
    "    \"mean_auctions_last_48hs\",\n",
    "    \"std_auctions_last_24hs\",\n",
    "    \"std_auctions_last_48hs\", \n",
    "    #\"mean_auctions_last_hour\", \n",
    "    #\"mean_auctions_last_4hs\", \n",
    "    \"mean_events_last_24hs\", \n",
    "    \"mean_events_last_48hs\",  \n",
    "    \"std_events_last_24hs\", \n",
    "    \"std_events_last_48hs\",\n",
    "    #\"mean_auctions_last_60hs\",\n",
    "    #\"mean_auctions_last_36hs\",\n",
    "    #\"mean_installs_last_24hs\",\n",
    "    #\"mean_installs_last_48hs\",\n",
    "    \"installs_per_auctions\"\n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [],
   "source": [
    "features_list_class_auc = [\n",
    "    'appearances_in_auctions', \n",
    "    'user_appeared_last_day', \n",
    "    'time_to_reappear',\n",
    "    'amount_of_clicks', \n",
    "    'has_installed', \n",
    "    'user_clicked_last_day',\n",
    "    'user_installed_last_day', \n",
    "    'amount_of_installs', \n",
    "    'cant_max_day',\n",
    "    'cant_min_day', \n",
    "    'mean_time_to_click', \n",
    "    'max_time_click',\n",
    "    'min_time_click', \n",
    "    'mean_auctions_per_day', \n",
    "    'mean_events_per_day',\n",
    "    'mean_clicks_per_day', \n",
    "    'amount_auctions_in_last_hour',\n",
    "    'amount_auctions_in_last_2_hours', \n",
    "    'amount_auctions_in_last_5_hours',\n",
    "    'amount_auctions_in_last_12_hours', \n",
    "    'amount_auctions_in_last_24_hours',\n",
    "    'amount_events_in_last_hour', \n",
    "    'amount_events_in_last_2_hours',\n",
    "    'amount_auctions_in_first_hour', \n",
    "    'amount_auctions_in_first_3_hours',\n",
    "    'amount_auctions_in_first_5_hours', \n",
    "    'amount_auctions_in_first_12_hours',\n",
    "    'amount_events_in_first_hour', \n",
    "    'amount_events_in_first_5_hours',\n",
    "    'amount_events_in_first_12_hours', \n",
    "    'amount_clicks_in_last_2_hours',\n",
    "    'amount_clicks_in_last_4_hours', \n",
    "    'device_os', \n",
    "    'std_time_to_click',\n",
    "    'std_time_auctions', \n",
    "    'max_time_install', \n",
    "    'min_time_install',\n",
    "    'mean_time_install', \n",
    "    'std_time_install', \n",
    "    'max_time_events',\n",
    "    'min_time_events', \n",
    "    'mean_time_events', \n",
    "    'std_time_events',\n",
    "    'installs_per_events', \n",
    "    'installs_per_clicks', \n",
    "    'events_x_app_210',\n",
    "    'events_x_app_122', \n",
    "    'events_x_app_65', \n",
    "    'events_x_app_121',\n",
    "    'events_x_app_26', \n",
    "    'most_installed_apps_used', \n",
    "    'cant_apps_used',\n",
    "    'cant_events_atributed', \n",
    "    'has_events_atributed',\n",
    "    'has_events_ids_with_installs', \n",
    "    'has_events_ids_without_installs',\n",
    "    'cant_events_0_4', \n",
    "    'cant_events_4_8', \n",
    "    'cant_events_8_12',\n",
    "    'cant_events_12_16', \n",
    "    'cant_events_16_20', \n",
    "    'cant_events_20_24',\n",
    "    'cant_auctions_0_4', \n",
    "    'cant_auctions_4_8', \n",
    "    'cant_auctions_8_12',\n",
    "    'cant_auctions_12_16', \n",
    "    'cant_auctions_16_20', \n",
    "    'cant_auctions_20_24',\n",
    "    'implicit', \n",
    "    'latitude', \n",
    "    'longitude', \n",
    "    'clicked_in_last_5_minutes',\n",
    "    'clicked_with_wifi_in_last_3_hours', \n",
    "    'hour_install', \n",
    "    'hour_events',\n",
    "    'hour_clicks', \n",
    "    'hour_auctions',\n",
    "    'amount_auctions_in_last_half_hour',\n",
    "    'amount_auctions_in_last_15_minutes',\n",
    "    \"mean_appearances_events\",\n",
    "    \"std_appearances_events\",\n",
    "    \"std_appearances_auctions\",\n",
    "    'cant_devices',\n",
    "    \"mean_appearances_auctions\",\n",
    "    \"mean_of_first_auction\",\n",
    "    \"mean_of_first_event\",\n",
    "    \"time_to_first_appearance_in_auctions\",\n",
    "    \"mean_auctions_last_24hs\",\n",
    "    \"mean_auctions_last_48hs\",\n",
    "    \"std_auctions_last_24hs\",\n",
    "    \"std_auctions_last_48hs\", \n",
    "    \"mean_auctions_last_hour\", \n",
    "    \"mean_auctions_last_4hs\", \n",
    "    \"mean_events_last_24hs\", \n",
    "    \"mean_events_last_48hs\",  \n",
    "    \"std_events_last_24hs\", \n",
    "    \"std_events_last_48hs\",\n",
    "    \"mean_auctions_last_60hs\",\n",
    "    \"mean_auctions_last_36hs\",\n",
    "    \"mean_installs_last_24hs\",\n",
    "    \"mean_installs_last_48hs\",\n",
    "    \"installs_per_auctions\",\n",
    "    \"dayofweek_auctions\",\n",
    "    \"dayofweek_installs\"\n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "features_list_class_inst = [\n",
    "    'appearances_in_auctions', \n",
    "    'user_appeared_last_day', \n",
    "    'time_to_reappear',\n",
    "    #'amount_of_clicks', \n",
    "    'has_installed', \n",
    "    'user_clicked_last_day',\n",
    "    'user_installed_last_day', \n",
    "    'amount_of_installs', \n",
    "    #'cant_max_day',\n",
    "    #'cant_min_day', \n",
    "    #'mean_time_to_click', \n",
    "    #'max_time_click',\n",
    "    #'min_time_click', \n",
    "    'mean_auctions_per_day', \n",
    "    'mean_events_per_day',\n",
    "    #'mean_clicks_per_day', \n",
    "    'amount_auctions_in_last_hour',\n",
    "    'amount_auctions_in_last_2_hours', \n",
    "    'amount_auctions_in_last_5_hours',\n",
    "    'amount_auctions_in_last_12_hours', \n",
    "    'amount_auctions_in_last_24_hours',\n",
    "    'amount_events_in_last_hour', \n",
    "    'amount_events_in_last_2_hours',\n",
    "    'amount_auctions_in_first_hour', \n",
    "    'amount_auctions_in_first_3_hours',\n",
    "    'amount_auctions_in_first_5_hours', \n",
    "    'amount_auctions_in_first_12_hours',\n",
    "    'amount_events_in_first_hour', \n",
    "    'amount_events_in_first_5_hours',\n",
    "    'amount_events_in_first_12_hours', \n",
    "    #'amount_clicks_in_last_2_hours',\n",
    "    #'amount_clicks_in_last_4_hours', \n",
    "    'device_os', \n",
    "    #'std_time_to_click',\n",
    "    'std_time_auctions', \n",
    "    #'max_time_install', \n",
    "    #'min_time_install',\n",
    "    #'mean_time_install', \n",
    "    #'std_time_install', \n",
    "    #'max_time_events',\n",
    "    #'min_time_events', \n",
    "    'mean_time_events', \n",
    "    'std_time_events',\n",
    "    #'installs_per_events', \n",
    "    #'installs_per_clicks', \n",
    "    'events_x_app_210',\n",
    "    'events_x_app_122', \n",
    "    'events_x_app_65', \n",
    "    'events_x_app_121',\n",
    "    'events_x_app_26', \n",
    "    'most_installed_apps_used', \n",
    "    'cant_apps_used',\n",
    "    'cant_events_atributed', \n",
    "    'has_events_atributed',\n",
    "    #'has_events_ids_with_installs', \n",
    "    #'has_events_ids_without_installs',\n",
    "    'cant_events_0_4', \n",
    "    'cant_events_4_8', \n",
    "    'cant_events_8_12',\n",
    "    'cant_events_12_16', \n",
    "    'cant_events_16_20', \n",
    "    'cant_events_20_24',\n",
    "    'cant_auctions_0_4', \n",
    "    'cant_auctions_4_8', \n",
    "    'cant_auctions_8_12',\n",
    "    'cant_auctions_12_16', \n",
    "    'cant_auctions_16_20', \n",
    "    'cant_auctions_20_24',\n",
    "    #'implicit', \n",
    "    #'latitude', \n",
    "    #'longitude', \n",
    "    #'clicked_in_last_5_minutes',\n",
    "    'clicked_with_wifi_in_last_3_hours', \n",
    "    'hour_install', \n",
    "    'hour_events',\n",
    "    'hour_clicks', \n",
    "    'hour_auctions',\n",
    "    'cant_installs_0_4',\n",
    "    'cant_installs_4_8', \n",
    "    'cant_installs_8_12',\n",
    "    'cant_installs_12_16',\n",
    "    'cant_installs_16_20',\n",
    "    'cant_installs_20_24',\n",
    "    \"mean_appearances_clicks\",\n",
    "    \"mean_appearances_installs\",\n",
    "    'amount_auctions_in_last_half_hour',\n",
    "    'amount_auctions_in_last_15_minutes',\n",
    "    'cant_devices', \n",
    "    \"mean_appearances_auctions\",\n",
    "   \"mean_appearances_events\",\n",
    "   \"std_appearances_events\",\n",
    "    \"std_appearances_auctions\",\n",
    "    \"mean_of_first_auction\",\n",
    "    \"mean_of_first_event\",\n",
    "    \"time_to_first_appearance_in_auctions\",\n",
    "    \"mean_auctions_last_24hs\",\n",
    "    \"mean_auctions_last_48hs\",\n",
    "    \"std_auctions_last_24hs\",\n",
    "    \"std_auctions_last_48hs\",\n",
    "    \"mean_auctions_last_hour\", \n",
    "    \"mean_auctions_last_4hs\", \n",
    "    \"mean_events_last_24hs\", \n",
    "    \"mean_events_last_48hs\",  \n",
    "    \"std_events_last_24hs\", \n",
    "    \"std_events_last_48hs\",\n",
    "    \"mean_auctions_last_60hs\",\n",
    "    \"mean_auctions_last_36hs\",\n",
    "    \"mean_installs_last_24hs\",\n",
    "    \"mean_installs_last_48hs\",\n",
    "    \"installs_per_auctions\"\n",
    "    \n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [],
   "source": [
    "features_list_class_inst = [\n",
    "    'appearances_in_auctions', \n",
    "    'user_appeared_last_day', \n",
    "    'time_to_reappear',\n",
    "    'amount_of_clicks', \n",
    "    'has_installed', \n",
    "    'user_clicked_last_day',\n",
    "    'user_installed_last_day', \n",
    "    'amount_of_installs', \n",
    "    'cant_max_day',\n",
    "    'cant_min_day', \n",
    "    'mean_time_to_click', \n",
    "    'max_time_click',\n",
    "    'min_time_click', \n",
    "    'mean_auctions_per_day', \n",
    "    'mean_events_per_day',\n",
    "    'mean_clicks_per_day', \n",
    "    'amount_auctions_in_last_hour',\n",
    "    'amount_auctions_in_last_2_hours', \n",
    "    'amount_auctions_in_last_5_hours',\n",
    "    'amount_auctions_in_last_12_hours', \n",
    "    'amount_auctions_in_last_24_hours',\n",
    "    'amount_events_in_last_hour', \n",
    "    'amount_events_in_last_2_hours',\n",
    "    'amount_auctions_in_first_hour', \n",
    "    'amount_auctions_in_first_3_hours',\n",
    "    'amount_auctions_in_first_5_hours', \n",
    "    'amount_auctions_in_first_12_hours',\n",
    "    'amount_events_in_first_hour', \n",
    "    'amount_events_in_first_5_hours',\n",
    "    'amount_events_in_first_12_hours', \n",
    "    'amount_clicks_in_last_2_hours',\n",
    "    'amount_clicks_in_last_4_hours', \n",
    "    'device_os', \n",
    "    'std_time_to_click',\n",
    "    'std_time_auctions', \n",
    "    'max_time_install', \n",
    "    'min_time_install',\n",
    "    'mean_time_install', \n",
    "    'std_time_install', \n",
    "    'max_time_events',\n",
    "    'min_time_events', \n",
    "    'mean_time_events', \n",
    "    'std_time_events',\n",
    "    'installs_per_events', \n",
    "    'installs_per_clicks', \n",
    "    'events_x_app_210',\n",
    "    'events_x_app_122', \n",
    "    'events_x_app_65', \n",
    "    'events_x_app_121',\n",
    "    'events_x_app_26', \n",
    "    'most_installed_apps_used', \n",
    "    'cant_apps_used',\n",
    "    'cant_events_atributed', \n",
    "    'has_events_atributed',\n",
    "    'has_events_ids_with_installs', \n",
    "    'has_events_ids_without_installs',\n",
    "    'cant_events_0_4', \n",
    "    'cant_events_4_8', \n",
    "    'cant_events_8_12',\n",
    "    'cant_events_12_16', \n",
    "    'cant_events_16_20', \n",
    "    'cant_events_20_24',\n",
    "    'cant_auctions_0_4', \n",
    "    'cant_auctions_4_8', \n",
    "    'cant_auctions_8_12',\n",
    "    'cant_auctions_12_16', \n",
    "    'cant_auctions_16_20', \n",
    "    'cant_auctions_20_24',\n",
    "    'implicit', \n",
    "    'latitude', \n",
    "    'longitude', \n",
    "    'clicked_in_last_5_minutes',\n",
    "    'clicked_with_wifi_in_last_3_hours', \n",
    "    'hour_install', \n",
    "    'hour_events',\n",
    "    'hour_clicks', \n",
    "    'hour_auctions',\n",
    "    'cant_installs_0_4',\n",
    "    'cant_installs_4_8', \n",
    "    'cant_installs_8_12',\n",
    "    'cant_installs_12_16',\n",
    "    'cant_installs_16_20',\n",
    "    'cant_installs_20_24',\n",
    "    \"mean_appearances_clicks\",\n",
    "    \"mean_appearances_installs\",\n",
    "    'amount_auctions_in_last_half_hour',\n",
    "    'amount_auctions_in_last_15_minutes',\n",
    "    'cant_devices', \n",
    "    \"mean_appearances_auctions\",\n",
    "    \"mean_appearances_events\",\n",
    "    \"std_appearances_events\",\n",
    "    \"std_appearances_auctions\",\n",
    "    \"mean_of_first_auction\",\n",
    "    \"mean_of_first_event\",\n",
    "    \"time_to_first_appearance_in_auctions\",\n",
    "    \"mean_auctions_last_24hs\",\n",
    "    \"mean_auctions_last_48hs\",\n",
    "    \"std_auctions_last_24hs\",\n",
    "    \"std_auctions_last_48hs\",\n",
    "    \"mean_auctions_last_hour\", \n",
    "    \"mean_auctions_last_4hs\", \n",
    "    \"mean_events_last_24hs\", \n",
    "    \"mean_events_last_48hs\",  \n",
    "    \"std_events_last_24hs\", \n",
    "    \"std_events_last_48hs\",\n",
    "    \"mean_auctions_last_60hs\",\n",
    "    \"mean_auctions_last_36hs\",\n",
    "    \"mean_installs_last_24hs\",\n",
    "    \"mean_installs_last_48hs\",\n",
    "    \"installs_per_auctions\",\n",
    "    \"dayofweek_auctions\",\n",
    "    \"dayofweek_installs\"\n",
    "]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Funciones útiles para el entrenamento y el balanceo"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_df_to_train(windows, features, labels, feature_list):\n",
    "    df_list = []\n",
    "    for window in windows:\n",
    "        df = select_features(features[window], feature_list).join(labels[window], how=\"inner\")\n",
    "        df_list.append(df)\n",
    "    df_full = pd.concat(df_list)\n",
    "    df_full.reset_index(inplace=True, drop=True)\n",
    "    return df_full"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# factor = 1 100% balanced\n",
    "# factor = 0 not balanced\n",
    "# label_value label que mas aparece y se debe balancear\n",
    "def balance(df_full, label_name, factor, label_value = 1):\n",
    "    cant_values = df_full[label_name].value_counts()[label_value]\n",
    "    cant_no_values = len(df_full[label_name]) - cant_values\n",
    "    a_borrar = int((cant_values-cant_no_values)*factor)\n",
    "    index_to_drop = np.random.choice(df_full.loc[df_full[label_name] == label_value].index,a_borrar, replace=False)\n",
    "    return df_full.drop(index=index_to_drop)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_train_test_split(df_full, label_name, test_size=0.3):\n",
    "    y = df_full[label_name]\n",
    "    X_data = df_full.drop(label_name, axis=1)\n",
    "    return train_test_split(X_data, y, test_size=test_size, random_state=7)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Importancia de los features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.datasets import load_boston\n",
    "from sklearn.ensemble import RandomForestRegressor"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> ### ***Importancia de los features de auctions***"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_auctions_importance = get_df_to_train(windows, features, label_auc, features_list_class_auc)\n",
    "df_auctions_importance = df_auctions_importance.sample(10000)\n",
    "\n",
    "\n",
    "df_auctions_features = df_auctions_importance.drop('label_auc', axis=1)\n",
    "df_auctions_labels = df_auctions_importance['label_auc']\n",
    "\n",
    "X = df_auctions_features.fillna(0)\n",
    "Y = df_auctions_labels\n",
    "names = X.columns.tolist()\n",
    "rf_auc = RandomForestRegressor(n_estimators=100)\n",
    "rf_auc.fit(X, Y)\n",
    "#print(\"Features sorted by their score:\")\n",
    "#print(sorted(zip(map(lambda x: round(x, 4), rf.feature_importances_), names), reverse=True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "feature_importance_auctions = sorted(zip(map(lambda x: round(x, 4), rf_auc.feature_importances_), names))\n",
    "feature_auctions_plot = pd.DataFrame(feature_importance_auctions).set_index(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Ploteo\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plot = feature_auctions_plot.plot(kind=\"barh\", figsize=(14,18))\n",
    "plot.set_title(\"Importance of auctions features\", fontsize=25)\n",
    "plot.set_ylabel(\"Feature\", fontsize=20)\n",
    "plot.set_xlabel(\"Importance\", fontsize=20);\n",
    "plot.legend([\"Importance\"], loc=4)\n",
    "plot.tick_params(labelsize=13, which='major')\n",
    "plt.savefig(\"importance_auctions.png\", bbox_inches=\"tight\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> ### ***Importancia de los features de installs***"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_installs_importance = get_df_to_train(windows, features, label_inst, features_list_class_inst)\n",
    "df_installs_importance = df_installs_importance.sample(50000)\n",
    "\n",
    "\n",
    "df_installs_features = df_installs_importance.drop('label_inst', axis=1)\n",
    "df_installs_labels = df_installs_importance['label_inst']\n",
    "\n",
    "X = df_installs_features.fillna(0)\n",
    "Y = df_installs_labels\n",
    "names = X.columns.tolist()\n",
    "rf_inst = RandomForestRegressor(n_estimators=100)\n",
    "rf_inst.fit(X, Y)\n",
    "#print(\"Features sorted by their score:\")\n",
    "#print(sorted(zip(map(lambda x: round(x, 4), rf.feature_importances_), names), reverse=True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "feature_importance_inst = sorted(zip(map(lambda x: round(x, 4), rf_inst.feature_importances_), names))\n",
    "feature_inst_plot = pd.DataFrame(feature_importance_inst).set_index(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Ploteo\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plot = feature_inst_plot.plot(kind=\"barh\", figsize=(14,18))\n",
    "plot.set_title(\"Importance of installs features\", fontsize=25)\n",
    "plot.set_ylabel(\"Feature\", fontsize=20)\n",
    "plot.set_xlabel(\"Importance\", fontsize=20);\n",
    "plot.legend([\"Importance\"], loc=4)\n",
    "plot.tick_params(labelsize=13, which='major')\n",
    "plt.savefig(\"importance_installs.png\", bbox_inches=\"tight\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ***Modelos***"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# **Redes neuronales**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout\n",
    "from keras import metrics"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> ### ***Modelo de redes neuronales de predicción de auctions***"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_full_auctions = get_df_to_train(windows, features, label_auc, features_list_class_auc)\n",
    "df_full_auctions = df_full_auctions.fillna(0)\n",
    "\n",
    "# Filtro los tiempos máximos.\n",
    "df_full_auctions = df_full_auctions[df_full_auctions[\"label_auc\"] < 259200]\n",
    "\n",
    "# Separo train y test.\n",
    "X_train_auc, X_test_auc, y_train_auc, y_test_auc = get_train_test_split(df_full_auctions, \"label_auc\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Creo el modelo\n",
    "model_auc_neural = Sequential()\n",
    "\n",
    "# Capa de entrada\n",
    "model_auc_neural.add(Dense(10, input_dim=X_train_auc.shape[1], activation='relu'))\n",
    "# Dropout para controlar el overfit\n",
    "model_auc_neural.add(Dropout(0.2))\n",
    "\n",
    "# Capa con 40 neuronas\n",
    "model_auc_neural.add(Dense(20, activation='relu'))\n",
    "# Dropout para controlar el overfit\n",
    "#model_auc_neural.add(Dropout(0.2))\n",
    "\n",
    "# Capa con 60 neuronas\n",
    "model_auc_neural.add(Dense(40, activation='relu'))\n",
    "# Dropout para controlar el overfit\n",
    "#model_auc_neural.add(Dropout(0.2))\n",
    "\n",
    "# Salida del modelo\n",
    "model_auc_neural.add(Dense(1))\n",
    "\n",
    "# Compilo el modelo\n",
    "model_auc_neural.compile(optimizer ='adam', loss = 'mean_squared_error', metrics =[metrics.mse])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Veo el modelo\n",
    "model_auc_neural.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Entreno el modelo\n",
    "model_auc_neural.fit(X_train_auc, y_train_auc, validation_data=(X_test_auc,y_test_auc), epochs=5, batch_size=32)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Acá pones el valor que da el \"mean_squared_error\" del modelo\n",
    "MSE = 4360503664.0435\n",
    "\n",
    "print(f\"RMSE: {MSE**(1/2)}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> ### ***Modelo de redes neuronales de predicción de installs***"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_full_inst = get_df_to_train(windows, features, label_inst, features_list_class_inst)\n",
    "df_full_inst = df_full_inst.fillna(0)\n",
    "\n",
    "# Saco los tiempos máximos\n",
    "\n",
    "df_full_inst = df_full_inst[df_full_inst[\"label_inst\"] < 259200]\n",
    "#df_full_inst = balance(df_full_inst, \"label_inst\", 1, 259200)\n",
    "\n",
    "\n",
    "# Separo en train y test\n",
    "X_train_inst, X_test_inst, y_train_inst, y_test_inst = get_train_test_split(df_full_inst, \"label_inst\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Creo el modelo\n",
    "model_inst_neural = Sequential()\n",
    "\n",
    "# Capa de entrada\n",
    "model_inst_neural.add(Dense(10, input_dim=X_train_inst.shape[1], activation='relu'))\n",
    "# Dropout para controlar el overfit\n",
    "\n",
    "\n",
    "# Capa con 30 neuronas\n",
    "model_inst_neural.add(Dense(20, activation='relu'))\n",
    "# Dropout para controlar el overfit\n",
    "\n",
    "# Capa con 50 neuronas\n",
    "model_inst_neural.add(Dense(40, activation='relu'))\n",
    "model_inst_neural.add(Dense(80, activation='relu'))\n",
    "model_inst_neural.add(Dense(120, activation='relu'))\n",
    "model_inst_neural.add(Dense(160, activation='relu'))\n",
    "# Dropout para controlar el overfit\n",
    "\n",
    "# Salida del modelo\n",
    "model_inst_neural.add(Dense(1))\n",
    "\n",
    "# Compilo el modelo\n",
    "model_inst_neural.compile(optimizer ='adam', loss = 'mean_squared_error', metrics =[metrics.mse])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Veo el modelo\n",
    "model_inst_neural.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Entreno el modelo\n",
    "model_inst_neural.fit(X_train_inst, y_train_inst, validation_data=(X_test_inst,y_test_inst), epochs=50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Acá pones el valor que da el \"mean_squared_error\" del modelo\n",
    "MSE = 6272147912.2780\n",
    "\n",
    "print(f\"RMSE: {MSE**(1/2)}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# **XGBoost**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> ### ***Modelo de XGBoost de predicción de auctions***"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_full = get_df_to_train(windows, features, label_auc, features_list_class_auc)\n",
    "\n",
    "#df_full = df_full.sample(int(len(df_full)))\n",
    "\n",
    "#df_full = balance(df_full, \"label_auc\", 0, 259200)\n",
    "\n",
    "df_full = df_full.loc[df_full[\"label_auc\"] != 259200]\n",
    "\n",
    "X_train, X_test, y_train, y_test = get_train_test_split(df_full, \"label_auc\", 0.2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_auc = xgb.XGBRegressor(   \n",
    "    gamma=0.1, \n",
    "    learning_rate=0.1, \n",
    "    max_depth=3,  \n",
    "    n_estimators=500,\n",
    "    n_jobs=2,  \n",
    "    objective='reg:linear',   \n",
    "    silent=True\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%time\n",
    "model_auc.fit(X_train, y_train, eval_metric='rmse')\n",
    "#cv = xgb.cv(params, xg_train, 5000, nfold=n_folds, early_stopping_rounds=early_stopping, verbose_eval=1)\n",
    "prediction = model_auc.predict(X_test)\n",
    "rmse = np.sqrt(mean_squared_error(y_test, prediction))\n",
    "print(\"RMSE: %f\" % (rmse))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_auc.save_model('xgboost_auc.model')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> #### ***Mejor modelo de auctions***"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "RMSE: 63345.641573\n",
    "CPU times: user 25min 48s, sys: 8.65 s, total: 25min 57s\n",
    "Wall time: 26min 8s\n",
    "    \n",
    "model_auc = xgb.XGBRegressor(\n",
    "    base_score=0.5, \n",
    "    booster='gbtree', \n",
    "    colsample_bylevel=1,\n",
    "    colsample_bytree=1, \n",
    "    gamma=0.1, \n",
    "    learning_rate=0.1, \n",
    "    max_delta_step=0,\n",
    "    max_depth=4, \n",
    "    min_child_weight=1, \n",
    "    missing=None, \n",
    "    n_estimators=600,\n",
    "    n_jobs=1, \n",
    "    nthread=None, \n",
    "    objective='reg:linear', \n",
    "    random_state=0,\n",
    "    reg_alpha=0, \n",
    "    reg_lambda=1, \n",
    "    scale_pos_weight=1, \n",
    "    seed=None,\n",
    "    silent=True, \n",
    "    subsample=1\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "_ = plot_importance(model_auc, height=0.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from xgboost import plot_importance, plot_tree\n",
    "_ = plot_importance(model_auc, height=0.5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> ### ***Modelo de XGBoost de predicción de installs***"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_full_installs = get_df_to_train(windows, features, label_inst, features_list_class_inst)\n",
    "\n",
    "df_full_installs = df_full_installs.sample(int(len(df_full_installs)))\n",
    "\n",
    "#df_full_installs = balance(df_full_installs, \"label_inst\", 1, 259200)\n",
    "df_full_installs = df_full_installs[df_full_installs[\"label_inst\"] != 259200]\n",
    "\n",
    "X_train, X_test, y_train, y_test = get_train_test_split(df_full_installs, \"label_inst\", 0.2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_inst = xgb.XGBRegressor(\n",
    "    learning_rate = 0.1,\n",
    "    n_estimators=700,\n",
    "    max_dept=6,\n",
    "    gamma=0.1,\n",
    "    n_jobs=-1\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%time\n",
    "model_inst.fit(X_train, y_train, eval_metric='rmse', verbose=True)\n",
    "\n",
    "prediction = model_inst.predict(X_test)\n",
    "rmse = np.sqrt(mean_squared_error(y_test, prediction))\n",
    "print(\"RMSE: %f\" % (rmse))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_inst.save_model('xgboost_inst.model')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from xgboost import plot_importance, plot_tree\n",
    "_ = plot_importance(model_inst, height=0.9)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> #### ***Mejor modelo de installs***"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "WARNING: src/objective/regression_obj.cu:152: reg:linear is now deprecated in favor of reg:squarederror.\n",
    "RMSE: 76768.119164\n",
    "CPU times: user 5min 39s, sys: 2 s, total: 5min 41s\n",
    "Wall time: 5min 44s\n",
    "\n",
    "model_inst = xgb.XGBRegressor(\n",
    "    learning_rate = 0.1,\n",
    "    n_estimators=700,\n",
    "    max_dept=5,\n",
    "    gamma=0.1,\n",
    "    n_jobs=-1\n",
    ")\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# **LightGBM**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### LightGBM para auctions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_full_auctions = get_df_to_train(windows, features, label_auc, features_list_class_auc)\n",
    "df_full_auctions = df_full_auctions[df_full_auctions[\"label_auc\"] != 259200]\n",
    "#df_full_auctions = df_full_auctions.sample()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train_auc, X_test_auc, y_train_auc, y_test_auc = get_train_test_split(df_full_auctions, \"label_auc\", 0.2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/lightgbm/engine.py:118: UserWarning: Found `n_estimators` in params. Will use it instead of argument\n",
      "  warnings.warn(\"Found `{}` in params. Will use it instead of argument\".format(alias))\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1]\tvalid_0's l2: 4.80641e+09\tvalid_0's l1: 56388.5\n",
      "[2]\tvalid_0's l2: 4.7355e+09\tvalid_0's l1: 55917.2\n",
      "[3]\tvalid_0's l2: 4.67114e+09\tvalid_0's l1: 55481.7\n",
      "[4]\tvalid_0's l2: 4.61294e+09\tvalid_0's l1: 55080.3\n",
      "[5]\tvalid_0's l2: 4.56018e+09\tvalid_0's l1: 54709.7\n",
      "[6]\tvalid_0's l2: 4.51237e+09\tvalid_0's l1: 54367.2\n",
      "[7]\tvalid_0's l2: 4.46912e+09\tvalid_0's l1: 54050.7\n",
      "[8]\tvalid_0's l2: 4.4298e+09\tvalid_0's l1: 53757.9\n",
      "[9]\tvalid_0's l2: 4.39403e+09\tvalid_0's l1: 53486.9\n",
      "[10]\tvalid_0's l2: 4.36161e+09\tvalid_0's l1: 53235.7\n",
      "[11]\tvalid_0's l2: 4.33221e+09\tvalid_0's l1: 53002.3\n",
      "[12]\tvalid_0's l2: 4.30557e+09\tvalid_0's l1: 52788.4\n",
      "[13]\tvalid_0's l2: 4.28114e+09\tvalid_0's l1: 52588.4\n",
      "[14]\tvalid_0's l2: 4.25896e+09\tvalid_0's l1: 52402.2\n",
      "[15]\tvalid_0's l2: 4.23899e+09\tvalid_0's l1: 52230.4\n",
      "[16]\tvalid_0's l2: 4.22078e+09\tvalid_0's l1: 52070.7\n",
      "[17]\tvalid_0's l2: 4.20388e+09\tvalid_0's l1: 51919\n",
      "[18]\tvalid_0's l2: 4.18832e+09\tvalid_0's l1: 51775.3\n",
      "[19]\tvalid_0's l2: 4.17457e+09\tvalid_0's l1: 51646\n",
      "[20]\tvalid_0's l2: 4.16183e+09\tvalid_0's l1: 51521.5\n",
      "[21]\tvalid_0's l2: 4.15028e+09\tvalid_0's l1: 51405.6\n",
      "[22]\tvalid_0's l2: 4.13982e+09\tvalid_0's l1: 51297.6\n",
      "[23]\tvalid_0's l2: 4.13021e+09\tvalid_0's l1: 51196.3\n",
      "[24]\tvalid_0's l2: 4.12077e+09\tvalid_0's l1: 51095\n",
      "[25]\tvalid_0's l2: 4.11235e+09\tvalid_0's l1: 51000.4\n",
      "[26]\tvalid_0's l2: 4.10471e+09\tvalid_0's l1: 50911.4\n",
      "[27]\tvalid_0's l2: 4.09756e+09\tvalid_0's l1: 50826\n",
      "[28]\tvalid_0's l2: 4.09105e+09\tvalid_0's l1: 50745.9\n",
      "[29]\tvalid_0's l2: 4.08548e+09\tvalid_0's l1: 50675\n",
      "[30]\tvalid_0's l2: 4.08011e+09\tvalid_0's l1: 50603.5\n",
      "[31]\tvalid_0's l2: 4.0753e+09\tvalid_0's l1: 50540.5\n",
      "[32]\tvalid_0's l2: 4.0708e+09\tvalid_0's l1: 50480.4\n",
      "[33]\tvalid_0's l2: 4.06657e+09\tvalid_0's l1: 50420.3\n",
      "[34]\tvalid_0's l2: 4.06281e+09\tvalid_0's l1: 50366.2\n",
      "[35]\tvalid_0's l2: 4.05926e+09\tvalid_0's l1: 50313.3\n",
      "[36]\tvalid_0's l2: 4.05605e+09\tvalid_0's l1: 50263.6\n",
      "[37]\tvalid_0's l2: 4.05311e+09\tvalid_0's l1: 50216.3\n",
      "[38]\tvalid_0's l2: 4.05039e+09\tvalid_0's l1: 50171.2\n",
      "[39]\tvalid_0's l2: 4.04795e+09\tvalid_0's l1: 50128.7\n",
      "[40]\tvalid_0's l2: 4.04574e+09\tvalid_0's l1: 50089.3\n",
      "[41]\tvalid_0's l2: 4.04364e+09\tvalid_0's l1: 50051.8\n",
      "[42]\tvalid_0's l2: 4.04161e+09\tvalid_0's l1: 50014.1\n",
      "[43]\tvalid_0's l2: 4.03982e+09\tvalid_0's l1: 49979.7\n",
      "[44]\tvalid_0's l2: 4.03818e+09\tvalid_0's l1: 49947.5\n",
      "[45]\tvalid_0's l2: 4.03658e+09\tvalid_0's l1: 49916.3\n",
      "[46]\tvalid_0's l2: 4.03521e+09\tvalid_0's l1: 49885.9\n",
      "[47]\tvalid_0's l2: 4.03394e+09\tvalid_0's l1: 49858.1\n",
      "[48]\tvalid_0's l2: 4.03267e+09\tvalid_0's l1: 49831.2\n",
      "[49]\tvalid_0's l2: 4.03146e+09\tvalid_0's l1: 49804.5\n",
      "[50]\tvalid_0's l2: 4.03044e+09\tvalid_0's l1: 49780.2\n",
      "[51]\tvalid_0's l2: 4.02946e+09\tvalid_0's l1: 49757\n",
      "[52]\tvalid_0's l2: 4.02858e+09\tvalid_0's l1: 49735.3\n",
      "[53]\tvalid_0's l2: 4.02767e+09\tvalid_0's l1: 49713.7\n",
      "[54]\tvalid_0's l2: 4.02689e+09\tvalid_0's l1: 49694.1\n",
      "[55]\tvalid_0's l2: 4.02623e+09\tvalid_0's l1: 49676.5\n",
      "[56]\tvalid_0's l2: 4.02547e+09\tvalid_0's l1: 49658.5\n",
      "[57]\tvalid_0's l2: 4.02478e+09\tvalid_0's l1: 49641.4\n",
      "[58]\tvalid_0's l2: 4.02414e+09\tvalid_0's l1: 49626.2\n",
      "[59]\tvalid_0's l2: 4.02357e+09\tvalid_0's l1: 49610.9\n",
      "[60]\tvalid_0's l2: 4.02307e+09\tvalid_0's l1: 49596.6\n",
      "[61]\tvalid_0's l2: 4.02251e+09\tvalid_0's l1: 49582\n",
      "[62]\tvalid_0's l2: 4.02201e+09\tvalid_0's l1: 49569.8\n",
      "[63]\tvalid_0's l2: 4.0215e+09\tvalid_0's l1: 49557\n",
      "[64]\tvalid_0's l2: 4.02101e+09\tvalid_0's l1: 49545.5\n",
      "[65]\tvalid_0's l2: 4.02062e+09\tvalid_0's l1: 49534.5\n",
      "[66]\tvalid_0's l2: 4.02019e+09\tvalid_0's l1: 49522.9\n",
      "[67]\tvalid_0's l2: 4.01987e+09\tvalid_0's l1: 49513.6\n",
      "[68]\tvalid_0's l2: 4.01952e+09\tvalid_0's l1: 49503.1\n",
      "[69]\tvalid_0's l2: 4.01919e+09\tvalid_0's l1: 49493.8\n",
      "[70]\tvalid_0's l2: 4.0189e+09\tvalid_0's l1: 49484.9\n",
      "[71]\tvalid_0's l2: 4.01859e+09\tvalid_0's l1: 49476.2\n",
      "[72]\tvalid_0's l2: 4.0183e+09\tvalid_0's l1: 49468.5\n",
      "[73]\tvalid_0's l2: 4.01807e+09\tvalid_0's l1: 49462\n",
      "[74]\tvalid_0's l2: 4.01781e+09\tvalid_0's l1: 49454.9\n",
      "[75]\tvalid_0's l2: 4.01761e+09\tvalid_0's l1: 49448.3\n",
      "[76]\tvalid_0's l2: 4.01741e+09\tvalid_0's l1: 49443\n",
      "[77]\tvalid_0's l2: 4.0172e+09\tvalid_0's l1: 49436.3\n",
      "[78]\tvalid_0's l2: 4.01698e+09\tvalid_0's l1: 49429.7\n",
      "[79]\tvalid_0's l2: 4.01676e+09\tvalid_0's l1: 49423.3\n",
      "[80]\tvalid_0's l2: 4.01662e+09\tvalid_0's l1: 49418.1\n",
      "[81]\tvalid_0's l2: 4.01638e+09\tvalid_0's l1: 49412\n",
      "[82]\tvalid_0's l2: 4.01621e+09\tvalid_0's l1: 49406.2\n",
      "[83]\tvalid_0's l2: 4.01601e+09\tvalid_0's l1: 49401\n",
      "[84]\tvalid_0's l2: 4.01585e+09\tvalid_0's l1: 49396.2\n",
      "[85]\tvalid_0's l2: 4.01567e+09\tvalid_0's l1: 49391.2\n",
      "[86]\tvalid_0's l2: 4.01553e+09\tvalid_0's l1: 49387.7\n",
      "[87]\tvalid_0's l2: 4.01538e+09\tvalid_0's l1: 49383.5\n",
      "[88]\tvalid_0's l2: 4.01524e+09\tvalid_0's l1: 49379.7\n",
      "[89]\tvalid_0's l2: 4.01512e+09\tvalid_0's l1: 49376\n",
      "[90]\tvalid_0's l2: 4.015e+09\tvalid_0's l1: 49373.2\n",
      "[91]\tvalid_0's l2: 4.01488e+09\tvalid_0's l1: 49369.8\n",
      "[92]\tvalid_0's l2: 4.01477e+09\tvalid_0's l1: 49367.6\n",
      "[93]\tvalid_0's l2: 4.01465e+09\tvalid_0's l1: 49364.7\n",
      "[94]\tvalid_0's l2: 4.01454e+09\tvalid_0's l1: 49362.1\n",
      "[95]\tvalid_0's l2: 4.01443e+09\tvalid_0's l1: 49358.7\n",
      "[96]\tvalid_0's l2: 4.01429e+09\tvalid_0's l1: 49356.1\n",
      "[97]\tvalid_0's l2: 4.01421e+09\tvalid_0's l1: 49353.9\n",
      "[98]\tvalid_0's l2: 4.01416e+09\tvalid_0's l1: 49351.7\n",
      "[99]\tvalid_0's l2: 4.01407e+09\tvalid_0's l1: 49349.3\n",
      "[100]\tvalid_0's l2: 4.01398e+09\tvalid_0's l1: 49347.6\n",
      "[101]\tvalid_0's l2: 4.01386e+09\tvalid_0's l1: 49345.8\n",
      "[102]\tvalid_0's l2: 4.01375e+09\tvalid_0's l1: 49344.3\n",
      "[103]\tvalid_0's l2: 4.01365e+09\tvalid_0's l1: 49342.7\n",
      "[104]\tvalid_0's l2: 4.01353e+09\tvalid_0's l1: 49340.2\n",
      "[105]\tvalid_0's l2: 4.01346e+09\tvalid_0's l1: 49339.2\n",
      "[106]\tvalid_0's l2: 4.01337e+09\tvalid_0's l1: 49336.9\n",
      "[107]\tvalid_0's l2: 4.0133e+09\tvalid_0's l1: 49335.5\n",
      "[108]\tvalid_0's l2: 4.01325e+09\tvalid_0's l1: 49334.3\n",
      "[109]\tvalid_0's l2: 4.0132e+09\tvalid_0's l1: 49332.7\n",
      "[110]\tvalid_0's l2: 4.01315e+09\tvalid_0's l1: 49331.9\n",
      "[111]\tvalid_0's l2: 4.01308e+09\tvalid_0's l1: 49331\n",
      "[112]\tvalid_0's l2: 4.01299e+09\tvalid_0's l1: 49329.6\n",
      "[113]\tvalid_0's l2: 4.0129e+09\tvalid_0's l1: 49328.1\n",
      "[114]\tvalid_0's l2: 4.01285e+09\tvalid_0's l1: 49326.5\n",
      "[115]\tvalid_0's l2: 4.01276e+09\tvalid_0's l1: 49325.1\n",
      "[116]\tvalid_0's l2: 4.01268e+09\tvalid_0's l1: 49324.3\n",
      "[117]\tvalid_0's l2: 4.01263e+09\tvalid_0's l1: 49323.3\n",
      "[118]\tvalid_0's l2: 4.01258e+09\tvalid_0's l1: 49323.1\n",
      "[119]\tvalid_0's l2: 4.01251e+09\tvalid_0's l1: 49322.7\n",
      "[120]\tvalid_0's l2: 4.01246e+09\tvalid_0's l1: 49322\n",
      "[121]\tvalid_0's l2: 4.01237e+09\tvalid_0's l1: 49320.4\n",
      "[122]\tvalid_0's l2: 4.01227e+09\tvalid_0's l1: 49318.8\n",
      "[123]\tvalid_0's l2: 4.01221e+09\tvalid_0's l1: 49317.2\n",
      "[124]\tvalid_0's l2: 4.01213e+09\tvalid_0's l1: 49315.8\n",
      "[125]\tvalid_0's l2: 4.01206e+09\tvalid_0's l1: 49314.6\n",
      "[126]\tvalid_0's l2: 4.01198e+09\tvalid_0's l1: 49313.7\n",
      "[127]\tvalid_0's l2: 4.0119e+09\tvalid_0's l1: 49313.1\n",
      "[128]\tvalid_0's l2: 4.01179e+09\tvalid_0's l1: 49311.7\n",
      "[129]\tvalid_0's l2: 4.01175e+09\tvalid_0's l1: 49311.2\n",
      "[130]\tvalid_0's l2: 4.01167e+09\tvalid_0's l1: 49310.2\n",
      "[131]\tvalid_0's l2: 4.0116e+09\tvalid_0's l1: 49309.1\n",
      "[132]\tvalid_0's l2: 4.01158e+09\tvalid_0's l1: 49308.4\n",
      "[133]\tvalid_0's l2: 4.01152e+09\tvalid_0's l1: 49307.6\n",
      "[134]\tvalid_0's l2: 4.0115e+09\tvalid_0's l1: 49306.6\n",
      "[135]\tvalid_0's l2: 4.01145e+09\tvalid_0's l1: 49306\n",
      "[136]\tvalid_0's l2: 4.01139e+09\tvalid_0's l1: 49305.4\n",
      "[137]\tvalid_0's l2: 4.01134e+09\tvalid_0's l1: 49305.1\n",
      "[138]\tvalid_0's l2: 4.0113e+09\tvalid_0's l1: 49304.2\n",
      "[139]\tvalid_0's l2: 4.01129e+09\tvalid_0's l1: 49303.5\n",
      "[140]\tvalid_0's l2: 4.01122e+09\tvalid_0's l1: 49302.6\n",
      "[141]\tvalid_0's l2: 4.01117e+09\tvalid_0's l1: 49302.2\n",
      "[142]\tvalid_0's l2: 4.01112e+09\tvalid_0's l1: 49301.5\n",
      "[143]\tvalid_0's l2: 4.01107e+09\tvalid_0's l1: 49300.9\n",
      "[144]\tvalid_0's l2: 4.01103e+09\tvalid_0's l1: 49300.2\n",
      "[145]\tvalid_0's l2: 4.01098e+09\tvalid_0's l1: 49299.8\n",
      "[146]\tvalid_0's l2: 4.01097e+09\tvalid_0's l1: 49298.6\n",
      "[147]\tvalid_0's l2: 4.01096e+09\tvalid_0's l1: 49298\n",
      "[148]\tvalid_0's l2: 4.01096e+09\tvalid_0's l1: 49296.9\n",
      "[149]\tvalid_0's l2: 4.01097e+09\tvalid_0's l1: 49296\n",
      "[150]\tvalid_0's l2: 4.01097e+09\tvalid_0's l1: 49295.4\n",
      "[151]\tvalid_0's l2: 4.01094e+09\tvalid_0's l1: 49295.1\n",
      "[152]\tvalid_0's l2: 4.0109e+09\tvalid_0's l1: 49294.2\n",
      "[153]\tvalid_0's l2: 4.01084e+09\tvalid_0's l1: 49293.4\n",
      "[154]\tvalid_0's l2: 4.01081e+09\tvalid_0's l1: 49293.1\n",
      "[155]\tvalid_0's l2: 4.0108e+09\tvalid_0's l1: 49292.3\n",
      "[156]\tvalid_0's l2: 4.01075e+09\tvalid_0's l1: 49291\n",
      "[157]\tvalid_0's l2: 4.01073e+09\tvalid_0's l1: 49290\n",
      "[158]\tvalid_0's l2: 4.01071e+09\tvalid_0's l1: 49289.2\n",
      "[159]\tvalid_0's l2: 4.01068e+09\tvalid_0's l1: 49288.1\n",
      "[160]\tvalid_0's l2: 4.01065e+09\tvalid_0's l1: 49287.3\n",
      "[161]\tvalid_0's l2: 4.01065e+09\tvalid_0's l1: 49287.7\n",
      "[162]\tvalid_0's l2: 4.01065e+09\tvalid_0's l1: 49288.4\n",
      "[163]\tvalid_0's l2: 4.01065e+09\tvalid_0's l1: 49288.8\n",
      "[164]\tvalid_0's l2: 4.01065e+09\tvalid_0's l1: 49289.1\n",
      "[165]\tvalid_0's l2: 4.01065e+09\tvalid_0's l1: 49290\n",
      "[166]\tvalid_0's l2: 4.01063e+09\tvalid_0's l1: 49289.3\n",
      "[167]\tvalid_0's l2: 4.01061e+09\tvalid_0's l1: 49288.1\n",
      "[168]\tvalid_0's l2: 4.01058e+09\tvalid_0's l1: 49287.3\n",
      "[169]\tvalid_0's l2: 4.01055e+09\tvalid_0's l1: 49287.1\n",
      "[170]\tvalid_0's l2: 4.01052e+09\tvalid_0's l1: 49287.1\n",
      "[171]\tvalid_0's l2: 4.01048e+09\tvalid_0's l1: 49286.5\n",
      "[172]\tvalid_0's l2: 4.01043e+09\tvalid_0's l1: 49285.2\n",
      "[173]\tvalid_0's l2: 4.01037e+09\tvalid_0's l1: 49284.4\n",
      "[174]\tvalid_0's l2: 4.01035e+09\tvalid_0's l1: 49283.8\n",
      "[175]\tvalid_0's l2: 4.01033e+09\tvalid_0's l1: 49282.7\n",
      "[176]\tvalid_0's l2: 4.01029e+09\tvalid_0's l1: 49282.3\n",
      "[177]\tvalid_0's l2: 4.01022e+09\tvalid_0's l1: 49281.3\n",
      "[178]\tvalid_0's l2: 4.01019e+09\tvalid_0's l1: 49280.9\n",
      "[179]\tvalid_0's l2: 4.01015e+09\tvalid_0's l1: 49279.9\n",
      "[180]\tvalid_0's l2: 4.01012e+09\tvalid_0's l1: 49279.5\n",
      "[181]\tvalid_0's l2: 4.01007e+09\tvalid_0's l1: 49278.6\n",
      "[182]\tvalid_0's l2: 4.01003e+09\tvalid_0's l1: 49278.3\n",
      "[183]\tvalid_0's l2: 4.00999e+09\tvalid_0's l1: 49278.1\n",
      "[184]\tvalid_0's l2: 4.00992e+09\tvalid_0's l1: 49276.9\n",
      "[185]\tvalid_0's l2: 4.00989e+09\tvalid_0's l1: 49276.7\n",
      "[186]\tvalid_0's l2: 4.00991e+09\tvalid_0's l1: 49277.3\n",
      "[187]\tvalid_0's l2: 4.00989e+09\tvalid_0's l1: 49277.2\n",
      "[188]\tvalid_0's l2: 4.0099e+09\tvalid_0's l1: 49277.5\n",
      "[189]\tvalid_0's l2: 4.00988e+09\tvalid_0's l1: 49277.3\n",
      "[190]\tvalid_0's l2: 4.00986e+09\tvalid_0's l1: 49276.8\n",
      "[191]\tvalid_0's l2: 4.00985e+09\tvalid_0's l1: 49276.7\n",
      "[192]\tvalid_0's l2: 4.00985e+09\tvalid_0's l1: 49276\n",
      "[193]\tvalid_0's l2: 4.00984e+09\tvalid_0's l1: 49275.4\n",
      "[194]\tvalid_0's l2: 4.00984e+09\tvalid_0's l1: 49275.4\n",
      "[195]\tvalid_0's l2: 4.00983e+09\tvalid_0's l1: 49275.2\n",
      "[196]\tvalid_0's l2: 4.00978e+09\tvalid_0's l1: 49274.8\n",
      "[197]\tvalid_0's l2: 4.00979e+09\tvalid_0's l1: 49274.7\n",
      "[198]\tvalid_0's l2: 4.00974e+09\tvalid_0's l1: 49273.9\n",
      "[199]\tvalid_0's l2: 4.00973e+09\tvalid_0's l1: 49273.8\n",
      "[200]\tvalid_0's l2: 4.00972e+09\tvalid_0's l1: 49273.3\n",
      "[201]\tvalid_0's l2: 4.0097e+09\tvalid_0's l1: 49273.7\n",
      "[202]\tvalid_0's l2: 4.00969e+09\tvalid_0's l1: 49273.8\n",
      "[203]\tvalid_0's l2: 4.00966e+09\tvalid_0's l1: 49274.2\n",
      "[204]\tvalid_0's l2: 4.00965e+09\tvalid_0's l1: 49274.6\n",
      "[205]\tvalid_0's l2: 4.00965e+09\tvalid_0's l1: 49274.8\n",
      "[206]\tvalid_0's l2: 4.00965e+09\tvalid_0's l1: 49274.8\n",
      "[207]\tvalid_0's l2: 4.00966e+09\tvalid_0's l1: 49274.6\n",
      "[208]\tvalid_0's l2: 4.00965e+09\tvalid_0's l1: 49275\n",
      "[209]\tvalid_0's l2: 4.00963e+09\tvalid_0's l1: 49274.9\n",
      "[210]\tvalid_0's l2: 4.00963e+09\tvalid_0's l1: 49275\n",
      "[211]\tvalid_0's l2: 4.00962e+09\tvalid_0's l1: 49275.5\n",
      "[212]\tvalid_0's l2: 4.0096e+09\tvalid_0's l1: 49276.2\n",
      "[213]\tvalid_0's l2: 4.0096e+09\tvalid_0's l1: 49276.4\n",
      "[214]\tvalid_0's l2: 4.00959e+09\tvalid_0's l1: 49276.7\n",
      "[215]\tvalid_0's l2: 4.00956e+09\tvalid_0's l1: 49277.1\n",
      "[216]\tvalid_0's l2: 4.00956e+09\tvalid_0's l1: 49276.6\n",
      "[217]\tvalid_0's l2: 4.00957e+09\tvalid_0's l1: 49275.8\n",
      "[218]\tvalid_0's l2: 4.00955e+09\tvalid_0's l1: 49275\n",
      "[219]\tvalid_0's l2: 4.00954e+09\tvalid_0's l1: 49273.9\n",
      "[220]\tvalid_0's l2: 4.00952e+09\tvalid_0's l1: 49272.8\n",
      "[221]\tvalid_0's l2: 4.00951e+09\tvalid_0's l1: 49272.8\n",
      "[222]\tvalid_0's l2: 4.00952e+09\tvalid_0's l1: 49272.6\n",
      "[223]\tvalid_0's l2: 4.00952e+09\tvalid_0's l1: 49272.9\n",
      "[224]\tvalid_0's l2: 4.0095e+09\tvalid_0's l1: 49272.6\n",
      "[225]\tvalid_0's l2: 4.00948e+09\tvalid_0's l1: 49272.1\n",
      "[226]\tvalid_0's l2: 4.00946e+09\tvalid_0's l1: 49271.4\n",
      "[227]\tvalid_0's l2: 4.00947e+09\tvalid_0's l1: 49271.1\n",
      "[228]\tvalid_0's l2: 4.00946e+09\tvalid_0's l1: 49270.9\n",
      "[229]\tvalid_0's l2: 4.00945e+09\tvalid_0's l1: 49270.4\n",
      "[230]\tvalid_0's l2: 4.00946e+09\tvalid_0's l1: 49269.7\n",
      "[231]\tvalid_0's l2: 4.00946e+09\tvalid_0's l1: 49269.1\n",
      "[232]\tvalid_0's l2: 4.00946e+09\tvalid_0's l1: 49268.6\n",
      "[233]\tvalid_0's l2: 4.00942e+09\tvalid_0's l1: 49267.8\n",
      "[234]\tvalid_0's l2: 4.00941e+09\tvalid_0's l1: 49267.2\n",
      "[235]\tvalid_0's l2: 4.00939e+09\tvalid_0's l1: 49266.2\n",
      "[236]\tvalid_0's l2: 4.00939e+09\tvalid_0's l1: 49266.3\n",
      "[237]\tvalid_0's l2: 4.00938e+09\tvalid_0's l1: 49265.9\n",
      "[238]\tvalid_0's l2: 4.00938e+09\tvalid_0's l1: 49265.8\n",
      "[239]\tvalid_0's l2: 4.00936e+09\tvalid_0's l1: 49265.4\n",
      "[240]\tvalid_0's l2: 4.00936e+09\tvalid_0's l1: 49265.4\n",
      "[241]\tvalid_0's l2: 4.00934e+09\tvalid_0's l1: 49265.3\n",
      "[242]\tvalid_0's l2: 4.00932e+09\tvalid_0's l1: 49265.3\n",
      "[243]\tvalid_0's l2: 4.00929e+09\tvalid_0's l1: 49265\n",
      "[244]\tvalid_0's l2: 4.00928e+09\tvalid_0's l1: 49264.9\n",
      "[245]\tvalid_0's l2: 4.00927e+09\tvalid_0's l1: 49264.8\n",
      "[246]\tvalid_0's l2: 4.00927e+09\tvalid_0's l1: 49264.9\n",
      "[247]\tvalid_0's l2: 4.00927e+09\tvalid_0's l1: 49265.4\n",
      "[248]\tvalid_0's l2: 4.00925e+09\tvalid_0's l1: 49264.8\n",
      "[249]\tvalid_0's l2: 4.00925e+09\tvalid_0's l1: 49264.9\n",
      "[250]\tvalid_0's l2: 4.00924e+09\tvalid_0's l1: 49264.9\n",
      "[251]\tvalid_0's l2: 4.00924e+09\tvalid_0's l1: 49264.6\n",
      "[252]\tvalid_0's l2: 4.00923e+09\tvalid_0's l1: 49264.4\n",
      "[253]\tvalid_0's l2: 4.00921e+09\tvalid_0's l1: 49264.4\n",
      "[254]\tvalid_0's l2: 4.00922e+09\tvalid_0's l1: 49264.4\n",
      "[255]\tvalid_0's l2: 4.00923e+09\tvalid_0's l1: 49264.3\n",
      "[256]\tvalid_0's l2: 4.00922e+09\tvalid_0's l1: 49264.2\n",
      "[257]\tvalid_0's l2: 4.00921e+09\tvalid_0's l1: 49263.9\n",
      "[258]\tvalid_0's l2: 4.00919e+09\tvalid_0's l1: 49263.8\n",
      "[259]\tvalid_0's l2: 4.00917e+09\tvalid_0's l1: 49263\n",
      "[260]\tvalid_0's l2: 4.00917e+09\tvalid_0's l1: 49262.4\n",
      "[261]\tvalid_0's l2: 4.00918e+09\tvalid_0's l1: 49262.8\n",
      "[262]\tvalid_0's l2: 4.00917e+09\tvalid_0's l1: 49262.9\n",
      "[263]\tvalid_0's l2: 4.00917e+09\tvalid_0's l1: 49263.1\n",
      "[264]\tvalid_0's l2: 4.00916e+09\tvalid_0's l1: 49263.2\n",
      "[265]\tvalid_0's l2: 4.00915e+09\tvalid_0's l1: 49263.1\n",
      "[266]\tvalid_0's l2: 4.00915e+09\tvalid_0's l1: 49263.1\n",
      "[267]\tvalid_0's l2: 4.00913e+09\tvalid_0's l1: 49263.6\n",
      "[268]\tvalid_0's l2: 4.00912e+09\tvalid_0's l1: 49263.9\n",
      "[269]\tvalid_0's l2: 4.00912e+09\tvalid_0's l1: 49264.3\n",
      "[270]\tvalid_0's l2: 4.00911e+09\tvalid_0's l1: 49264.9\n",
      "[271]\tvalid_0's l2: 4.00911e+09\tvalid_0's l1: 49264.8\n",
      "[272]\tvalid_0's l2: 4.0091e+09\tvalid_0's l1: 49265.2\n",
      "[273]\tvalid_0's l2: 4.00909e+09\tvalid_0's l1: 49264.9\n",
      "[274]\tvalid_0's l2: 4.00905e+09\tvalid_0's l1: 49264.9\n",
      "[275]\tvalid_0's l2: 4.00906e+09\tvalid_0's l1: 49264.7\n",
      "[276]\tvalid_0's l2: 4.00906e+09\tvalid_0's l1: 49264\n",
      "[277]\tvalid_0's l2: 4.00906e+09\tvalid_0's l1: 49262.7\n",
      "[278]\tvalid_0's l2: 4.00906e+09\tvalid_0's l1: 49262.3\n",
      "[279]\tvalid_0's l2: 4.00906e+09\tvalid_0's l1: 49261.8\n",
      "[280]\tvalid_0's l2: 4.00908e+09\tvalid_0's l1: 49261.2\n",
      "[281]\tvalid_0's l2: 4.00908e+09\tvalid_0's l1: 49262.1\n",
      "[282]\tvalid_0's l2: 4.00908e+09\tvalid_0's l1: 49262.8\n",
      "[283]\tvalid_0's l2: 4.00908e+09\tvalid_0's l1: 49263.1\n",
      "[284]\tvalid_0's l2: 4.00908e+09\tvalid_0's l1: 49263.8\n",
      "[285]\tvalid_0's l2: 4.00908e+09\tvalid_0's l1: 49264.8\n",
      "[286]\tvalid_0's l2: 4.00907e+09\tvalid_0's l1: 49264.2\n",
      "[287]\tvalid_0's l2: 4.00905e+09\tvalid_0's l1: 49263.7\n",
      "[288]\tvalid_0's l2: 4.00903e+09\tvalid_0's l1: 49263.1\n",
      "[289]\tvalid_0's l2: 4.00902e+09\tvalid_0's l1: 49262.9\n",
      "[290]\tvalid_0's l2: 4.00903e+09\tvalid_0's l1: 49262.6\n",
      "[291]\tvalid_0's l2: 4.00904e+09\tvalid_0's l1: 49262.5\n",
      "[292]\tvalid_0's l2: 4.00901e+09\tvalid_0's l1: 49262.3\n",
      "[293]\tvalid_0's l2: 4.00901e+09\tvalid_0's l1: 49262.1\n",
      "[294]\tvalid_0's l2: 4.009e+09\tvalid_0's l1: 49261.9\n",
      "[295]\tvalid_0's l2: 4.00902e+09\tvalid_0's l1: 49262.2\n",
      "[296]\tvalid_0's l2: 4.00901e+09\tvalid_0's l1: 49262.4\n",
      "[297]\tvalid_0's l2: 4.00899e+09\tvalid_0's l1: 49262.9\n",
      "[298]\tvalid_0's l2: 4.009e+09\tvalid_0's l1: 49262.9\n",
      "[299]\tvalid_0's l2: 4.009e+09\tvalid_0's l1: 49263.2\n",
      "[300]\tvalid_0's l2: 4.00899e+09\tvalid_0's l1: 49263.7\n",
      "[301]\tvalid_0's l2: 4.00898e+09\tvalid_0's l1: 49263.2\n",
      "[302]\tvalid_0's l2: 4.00896e+09\tvalid_0's l1: 49263\n",
      "[303]\tvalid_0's l2: 4.00895e+09\tvalid_0's l1: 49262.7\n",
      "[304]\tvalid_0's l2: 4.00892e+09\tvalid_0's l1: 49262.8\n",
      "[305]\tvalid_0's l2: 4.00893e+09\tvalid_0's l1: 49262.8\n",
      "[306]\tvalid_0's l2: 4.0089e+09\tvalid_0's l1: 49262.5\n",
      "[307]\tvalid_0's l2: 4.00892e+09\tvalid_0's l1: 49262.2\n",
      "[308]\tvalid_0's l2: 4.00893e+09\tvalid_0's l1: 49261.9\n",
      "[309]\tvalid_0's l2: 4.00893e+09\tvalid_0's l1: 49261.5\n",
      "[310]\tvalid_0's l2: 4.00894e+09\tvalid_0's l1: 49261.2\n",
      "[311]\tvalid_0's l2: 4.00895e+09\tvalid_0's l1: 49261.4\n",
      "[312]\tvalid_0's l2: 4.00894e+09\tvalid_0's l1: 49261.3\n",
      "[313]\tvalid_0's l2: 4.00893e+09\tvalid_0's l1: 49261.5\n",
      "[314]\tvalid_0's l2: 4.00894e+09\tvalid_0's l1: 49261.4\n",
      "[315]\tvalid_0's l2: 4.00893e+09\tvalid_0's l1: 49261.7\n",
      "[316]\tvalid_0's l2: 4.00891e+09\tvalid_0's l1: 49261\n",
      "[317]\tvalid_0's l2: 4.00892e+09\tvalid_0's l1: 49260.1\n",
      "[318]\tvalid_0's l2: 4.00889e+09\tvalid_0's l1: 49259.5\n",
      "[319]\tvalid_0's l2: 4.00886e+09\tvalid_0's l1: 49258.5\n",
      "[320]\tvalid_0's l2: 4.00886e+09\tvalid_0's l1: 49258.2\n",
      "[321]\tvalid_0's l2: 4.00886e+09\tvalid_0's l1: 49258.2\n",
      "[322]\tvalid_0's l2: 4.00885e+09\tvalid_0's l1: 49258\n",
      "[323]\tvalid_0's l2: 4.00885e+09\tvalid_0's l1: 49258.4\n",
      "[324]\tvalid_0's l2: 4.00885e+09\tvalid_0's l1: 49258.4\n",
      "[325]\tvalid_0's l2: 4.00885e+09\tvalid_0's l1: 49258.7\n",
      "[326]\tvalid_0's l2: 4.00884e+09\tvalid_0's l1: 49258.4\n",
      "[327]\tvalid_0's l2: 4.00886e+09\tvalid_0's l1: 49258.1\n",
      "[328]\tvalid_0's l2: 4.00886e+09\tvalid_0's l1: 49257.6\n",
      "[329]\tvalid_0's l2: 4.00886e+09\tvalid_0's l1: 49257.2\n",
      "[330]\tvalid_0's l2: 4.00887e+09\tvalid_0's l1: 49256.9\n",
      "[331]\tvalid_0's l2: 4.00887e+09\tvalid_0's l1: 49256.8\n",
      "[332]\tvalid_0's l2: 4.00887e+09\tvalid_0's l1: 49257.1\n",
      "[333]\tvalid_0's l2: 4.00886e+09\tvalid_0's l1: 49256.8\n",
      "[334]\tvalid_0's l2: 4.00886e+09\tvalid_0's l1: 49256\n",
      "[335]\tvalid_0's l2: 4.00888e+09\tvalid_0's l1: 49255.1\n",
      "[336]\tvalid_0's l2: 4.0089e+09\tvalid_0's l1: 49254.3\n",
      "[337]\tvalid_0's l2: 4.0089e+09\tvalid_0's l1: 49253.9\n",
      "[338]\tvalid_0's l2: 4.0089e+09\tvalid_0's l1: 49253.4\n",
      "[339]\tvalid_0's l2: 4.00891e+09\tvalid_0's l1: 49252.8\n",
      "[340]\tvalid_0's l2: 4.00892e+09\tvalid_0's l1: 49252.6\n",
      "[341]\tvalid_0's l2: 4.00891e+09\tvalid_0's l1: 49252.6\n",
      "[342]\tvalid_0's l2: 4.00895e+09\tvalid_0's l1: 49252.8\n",
      "[343]\tvalid_0's l2: 4.00895e+09\tvalid_0's l1: 49252.3\n",
      "[344]\tvalid_0's l2: 4.00895e+09\tvalid_0's l1: 49251.8\n",
      "[345]\tvalid_0's l2: 4.00896e+09\tvalid_0's l1: 49251.3\n",
      "[346]\tvalid_0's l2: 4.00896e+09\tvalid_0's l1: 49251.2\n",
      "[347]\tvalid_0's l2: 4.00892e+09\tvalid_0's l1: 49250.8\n",
      "[348]\tvalid_0's l2: 4.00893e+09\tvalid_0's l1: 49251\n",
      "[349]\tvalid_0's l2: 4.00894e+09\tvalid_0's l1: 49251\n",
      "[350]\tvalid_0's l2: 4.00892e+09\tvalid_0's l1: 49250.7\n",
      "[351]\tvalid_0's l2: 4.00892e+09\tvalid_0's l1: 49250.5\n",
      "[352]\tvalid_0's l2: 4.00892e+09\tvalid_0's l1: 49250\n",
      "[353]\tvalid_0's l2: 4.00892e+09\tvalid_0's l1: 49250\n",
      "[354]\tvalid_0's l2: 4.00894e+09\tvalid_0's l1: 49249.8\n",
      "[355]\tvalid_0's l2: 4.00893e+09\tvalid_0's l1: 49249.6\n",
      "[356]\tvalid_0's l2: 4.00894e+09\tvalid_0's l1: 49249.7\n",
      "[357]\tvalid_0's l2: 4.00893e+09\tvalid_0's l1: 49249.6\n",
      "[358]\tvalid_0's l2: 4.00894e+09\tvalid_0's l1: 49249.6\n",
      "[359]\tvalid_0's l2: 4.00892e+09\tvalid_0's l1: 49249.3\n",
      "[360]\tvalid_0's l2: 4.00893e+09\tvalid_0's l1: 49249.4\n",
      "[361]\tvalid_0's l2: 4.00889e+09\tvalid_0's l1: 49249.3\n",
      "[362]\tvalid_0's l2: 4.00888e+09\tvalid_0's l1: 49249\n",
      "[363]\tvalid_0's l2: 4.00887e+09\tvalid_0's l1: 49248.9\n",
      "[364]\tvalid_0's l2: 4.00887e+09\tvalid_0's l1: 49248.3\n",
      "[365]\tvalid_0's l2: 4.00889e+09\tvalid_0's l1: 49248.1\n",
      "[366]\tvalid_0's l2: 4.00891e+09\tvalid_0's l1: 49248.9\n",
      "[367]\tvalid_0's l2: 4.00891e+09\tvalid_0's l1: 49249.6\n",
      "[368]\tvalid_0's l2: 4.00888e+09\tvalid_0's l1: 49249.9\n",
      "[369]\tvalid_0's l2: 4.00887e+09\tvalid_0's l1: 49250.5\n",
      "[370]\tvalid_0's l2: 4.00889e+09\tvalid_0's l1: 49251.2\n",
      "[371]\tvalid_0's l2: 4.00891e+09\tvalid_0's l1: 49251.8\n",
      "[372]\tvalid_0's l2: 4.0089e+09\tvalid_0's l1: 49251.9\n",
      "[373]\tvalid_0's l2: 4.00891e+09\tvalid_0's l1: 49252.3\n",
      "[374]\tvalid_0's l2: 4.00891e+09\tvalid_0's l1: 49252.4\n",
      "[375]\tvalid_0's l2: 4.00891e+09\tvalid_0's l1: 49252.5\n",
      "[376]\tvalid_0's l2: 4.00892e+09\tvalid_0's l1: 49253.2\n",
      "[377]\tvalid_0's l2: 4.00891e+09\tvalid_0's l1: 49253\n",
      "[378]\tvalid_0's l2: 4.00891e+09\tvalid_0's l1: 49253.3\n",
      "[379]\tvalid_0's l2: 4.00889e+09\tvalid_0's l1: 49253.6\n",
      "[380]\tvalid_0's l2: 4.00888e+09\tvalid_0's l1: 49253.8\n",
      "[381]\tvalid_0's l2: 4.00887e+09\tvalid_0's l1: 49254\n",
      "[382]\tvalid_0's l2: 4.00886e+09\tvalid_0's l1: 49254.1\n",
      "[383]\tvalid_0's l2: 4.00884e+09\tvalid_0's l1: 49254.6\n",
      "[384]\tvalid_0's l2: 4.00885e+09\tvalid_0's l1: 49254.7\n",
      "[385]\tvalid_0's l2: 4.00884e+09\tvalid_0's l1: 49254.7\n",
      "[386]\tvalid_0's l2: 4.00884e+09\tvalid_0's l1: 49255\n",
      "[387]\tvalid_0's l2: 4.00884e+09\tvalid_0's l1: 49255.2\n",
      "[388]\tvalid_0's l2: 4.00886e+09\tvalid_0's l1: 49255.6\n",
      "[389]\tvalid_0's l2: 4.00888e+09\tvalid_0's l1: 49256.4\n",
      "[390]\tvalid_0's l2: 4.00887e+09\tvalid_0's l1: 49256.6\n",
      "[391]\tvalid_0's l2: 4.00887e+09\tvalid_0's l1: 49256.2\n",
      "[392]\tvalid_0's l2: 4.00887e+09\tvalid_0's l1: 49255.7\n",
      "[393]\tvalid_0's l2: 4.00888e+09\tvalid_0's l1: 49255.1\n",
      "[394]\tvalid_0's l2: 4.00889e+09\tvalid_0's l1: 49254.5\n",
      "[395]\tvalid_0's l2: 4.00888e+09\tvalid_0's l1: 49254.6\n",
      "[396]\tvalid_0's l2: 4.00887e+09\tvalid_0's l1: 49254.3\n",
      "[397]\tvalid_0's l2: 4.00887e+09\tvalid_0's l1: 49254.5\n",
      "[398]\tvalid_0's l2: 4.00885e+09\tvalid_0's l1: 49254.1\n",
      "[399]\tvalid_0's l2: 4.00885e+09\tvalid_0's l1: 49253.9\n",
      "[400]\tvalid_0's l2: 4.00887e+09\tvalid_0's l1: 49254\n",
      "[401]\tvalid_0's l2: 4.00887e+09\tvalid_0's l1: 49253.4\n",
      "[402]\tvalid_0's l2: 4.00885e+09\tvalid_0's l1: 49253.5\n",
      "[403]\tvalid_0's l2: 4.00882e+09\tvalid_0's l1: 49253.7\n",
      "[404]\tvalid_0's l2: 4.00883e+09\tvalid_0's l1: 49253.4\n",
      "[405]\tvalid_0's l2: 4.00882e+09\tvalid_0's l1: 49253.3\n",
      "[406]\tvalid_0's l2: 4.00883e+09\tvalid_0's l1: 49253.5\n",
      "[407]\tvalid_0's l2: 4.00882e+09\tvalid_0's l1: 49253.7\n",
      "[408]\tvalid_0's l2: 4.00882e+09\tvalid_0's l1: 49254.1\n",
      "[409]\tvalid_0's l2: 4.00881e+09\tvalid_0's l1: 49254.4\n",
      "[410]\tvalid_0's l2: 4.00883e+09\tvalid_0's l1: 49254.8\n",
      "[411]\tvalid_0's l2: 4.00883e+09\tvalid_0's l1: 49254.9\n",
      "[412]\tvalid_0's l2: 4.00883e+09\tvalid_0's l1: 49255\n",
      "[413]\tvalid_0's l2: 4.00886e+09\tvalid_0's l1: 49255.3\n",
      "[414]\tvalid_0's l2: 4.00887e+09\tvalid_0's l1: 49255.6\n",
      "[415]\tvalid_0's l2: 4.00886e+09\tvalid_0's l1: 49255.5\n",
      "[416]\tvalid_0's l2: 4.00886e+09\tvalid_0's l1: 49254.9\n",
      "[417]\tvalid_0's l2: 4.00885e+09\tvalid_0's l1: 49254.9\n",
      "[418]\tvalid_0's l2: 4.00885e+09\tvalid_0's l1: 49254.7\n",
      "[419]\tvalid_0's l2: 4.00886e+09\tvalid_0's l1: 49254.5\n",
      "[420]\tvalid_0's l2: 4.00888e+09\tvalid_0's l1: 49254.2\n",
      "[421]\tvalid_0's l2: 4.00885e+09\tvalid_0's l1: 49253.9\n",
      "[422]\tvalid_0's l2: 4.00887e+09\tvalid_0's l1: 49254.1\n",
      "[423]\tvalid_0's l2: 4.00886e+09\tvalid_0's l1: 49254.4\n",
      "[424]\tvalid_0's l2: 4.00886e+09\tvalid_0's l1: 49254.2\n",
      "[425]\tvalid_0's l2: 4.00889e+09\tvalid_0's l1: 49254.2\n",
      "[426]\tvalid_0's l2: 4.00891e+09\tvalid_0's l1: 49253.5\n",
      "[427]\tvalid_0's l2: 4.0089e+09\tvalid_0's l1: 49252.8\n",
      "[428]\tvalid_0's l2: 4.00889e+09\tvalid_0's l1: 49251.8\n",
      "[429]\tvalid_0's l2: 4.00888e+09\tvalid_0's l1: 49251.3\n",
      "[430]\tvalid_0's l2: 4.00885e+09\tvalid_0's l1: 49250.7\n",
      "[431]\tvalid_0's l2: 4.00887e+09\tvalid_0's l1: 49250.9\n",
      "[432]\tvalid_0's l2: 4.00886e+09\tvalid_0's l1: 49251\n",
      "[433]\tvalid_0's l2: 4.00885e+09\tvalid_0's l1: 49251\n",
      "[434]\tvalid_0's l2: 4.00884e+09\tvalid_0's l1: 49250.6\n",
      "[435]\tvalid_0's l2: 4.00887e+09\tvalid_0's l1: 49250.8\n",
      "[436]\tvalid_0's l2: 4.00883e+09\tvalid_0's l1: 49250.5\n",
      "[437]\tvalid_0's l2: 4.0088e+09\tvalid_0's l1: 49250.4\n",
      "[438]\tvalid_0's l2: 4.00878e+09\tvalid_0's l1: 49250.5\n",
      "[439]\tvalid_0's l2: 4.00877e+09\tvalid_0's l1: 49250.5\n",
      "[440]\tvalid_0's l2: 4.00878e+09\tvalid_0's l1: 49250.6\n",
      "[441]\tvalid_0's l2: 4.00879e+09\tvalid_0's l1: 49250.5\n",
      "[442]\tvalid_0's l2: 4.00877e+09\tvalid_0's l1: 49250.2\n",
      "[443]\tvalid_0's l2: 4.00879e+09\tvalid_0's l1: 49250.3\n",
      "[444]\tvalid_0's l2: 4.00878e+09\tvalid_0's l1: 49250.2\n",
      "[445]\tvalid_0's l2: 4.00877e+09\tvalid_0's l1: 49250.1\n",
      "[446]\tvalid_0's l2: 4.00878e+09\tvalid_0's l1: 49250.8\n",
      "[447]\tvalid_0's l2: 4.00881e+09\tvalid_0's l1: 49251.3\n",
      "[448]\tvalid_0's l2: 4.00881e+09\tvalid_0's l1: 49251.1\n",
      "[449]\tvalid_0's l2: 4.00883e+09\tvalid_0's l1: 49251.4\n",
      "[450]\tvalid_0's l2: 4.00884e+09\tvalid_0's l1: 49251.5\n",
      "[451]\tvalid_0's l2: 4.00886e+09\tvalid_0's l1: 49251.5\n",
      "[452]\tvalid_0's l2: 4.00888e+09\tvalid_0's l1: 49251.3\n",
      "[453]\tvalid_0's l2: 4.0089e+09\tvalid_0's l1: 49251.4\n",
      "[454]\tvalid_0's l2: 4.00889e+09\tvalid_0's l1: 49251.1\n",
      "[455]\tvalid_0's l2: 4.0089e+09\tvalid_0's l1: 49250.6\n",
      "[456]\tvalid_0's l2: 4.00889e+09\tvalid_0's l1: 49250.6\n",
      "[457]\tvalid_0's l2: 4.0089e+09\tvalid_0's l1: 49250.7\n",
      "[458]\tvalid_0's l2: 4.00891e+09\tvalid_0's l1: 49250.8\n",
      "[459]\tvalid_0's l2: 4.00892e+09\tvalid_0's l1: 49251.1\n",
      "[460]\tvalid_0's l2: 4.00892e+09\tvalid_0's l1: 49251.7\n",
      "[461]\tvalid_0's l2: 4.00892e+09\tvalid_0's l1: 49251.5\n",
      "[462]\tvalid_0's l2: 4.00892e+09\tvalid_0's l1: 49251.3\n",
      "[463]\tvalid_0's l2: 4.00893e+09\tvalid_0's l1: 49251\n",
      "[464]\tvalid_0's l2: 4.00892e+09\tvalid_0's l1: 49251.1\n",
      "[465]\tvalid_0's l2: 4.00892e+09\tvalid_0's l1: 49250.6\n",
      "[466]\tvalid_0's l2: 4.00891e+09\tvalid_0's l1: 49250.7\n",
      "[467]\tvalid_0's l2: 4.0089e+09\tvalid_0's l1: 49250.6\n",
      "[468]\tvalid_0's l2: 4.0089e+09\tvalid_0's l1: 49250.7\n",
      "[469]\tvalid_0's l2: 4.00891e+09\tvalid_0's l1: 49250.6\n",
      "[470]\tvalid_0's l2: 4.00891e+09\tvalid_0's l1: 49250.6\n",
      "[471]\tvalid_0's l2: 4.0089e+09\tvalid_0's l1: 49251.1\n",
      "[472]\tvalid_0's l2: 4.00893e+09\tvalid_0's l1: 49251.4\n",
      "[473]\tvalid_0's l2: 4.00896e+09\tvalid_0's l1: 49251.7\n",
      "[474]\tvalid_0's l2: 4.00896e+09\tvalid_0's l1: 49252.2\n",
      "[475]\tvalid_0's l2: 4.00898e+09\tvalid_0's l1: 49252.6\n",
      "[476]\tvalid_0's l2: 4.00898e+09\tvalid_0's l1: 49253.3\n",
      "[477]\tvalid_0's l2: 4.00898e+09\tvalid_0's l1: 49254\n",
      "[478]\tvalid_0's l2: 4.00898e+09\tvalid_0's l1: 49254\n",
      "[479]\tvalid_0's l2: 4.00899e+09\tvalid_0's l1: 49254.4\n",
      "[480]\tvalid_0's l2: 4.00899e+09\tvalid_0's l1: 49254.9\n",
      "[481]\tvalid_0's l2: 4.00899e+09\tvalid_0's l1: 49254\n",
      "[482]\tvalid_0's l2: 4.00901e+09\tvalid_0's l1: 49254.4\n",
      "[483]\tvalid_0's l2: 4.00903e+09\tvalid_0's l1: 49254.5\n",
      "[484]\tvalid_0's l2: 4.00901e+09\tvalid_0's l1: 49254.3\n",
      "[485]\tvalid_0's l2: 4.009e+09\tvalid_0's l1: 49254.3\n",
      "[486]\tvalid_0's l2: 4.00901e+09\tvalid_0's l1: 49254\n",
      "[487]\tvalid_0's l2: 4.009e+09\tvalid_0's l1: 49253.8\n",
      "[488]\tvalid_0's l2: 4.00899e+09\tvalid_0's l1: 49253.4\n",
      "[489]\tvalid_0's l2: 4.00901e+09\tvalid_0's l1: 49253\n",
      "[490]\tvalid_0's l2: 4.009e+09\tvalid_0's l1: 49252.6\n",
      "[491]\tvalid_0's l2: 4.00899e+09\tvalid_0's l1: 49252.6\n",
      "[492]\tvalid_0's l2: 4.009e+09\tvalid_0's l1: 49253.2\n",
      "[493]\tvalid_0's l2: 4.00898e+09\tvalid_0's l1: 49253.2\n",
      "[494]\tvalid_0's l2: 4.00898e+09\tvalid_0's l1: 49253.3\n",
      "[495]\tvalid_0's l2: 4.00896e+09\tvalid_0's l1: 49253.1\n",
      "[496]\tvalid_0's l2: 4.00895e+09\tvalid_0's l1: 49253.1\n",
      "[497]\tvalid_0's l2: 4.00895e+09\tvalid_0's l1: 49253.3\n",
      "[498]\tvalid_0's l2: 4.00893e+09\tvalid_0's l1: 49252.9\n",
      "[499]\tvalid_0's l2: 4.00893e+09\tvalid_0's l1: 49253.1\n",
      "[500]\tvalid_0's l2: 4.00891e+09\tvalid_0's l1: 49253.1\n",
      "The rmse of prediction is: 63315.97962183445\n"
     ]
    }
   ],
   "source": [
    "# create dataset for lightgbm\n",
    "lgb_train = lgb.Dataset(X_train_auc, y_train_auc)\n",
    "lgb_eval = lgb.Dataset(X_test_auc, y_test_auc, reference=lgb_train)\n",
    "\n",
    "params = {\n",
    "    'n_estimators' : 500,\n",
    "    'boosting_type': 'gbdt',\n",
    "    #'boosting_type': 'dart',\n",
    "    'objective': 'regression',\n",
    "    'metric': {'l1', 'l2'},\n",
    "    'learning_rate': 0.05,\n",
    "    'feature_fraction': 0.9,\n",
    "    'bagging_fraction': 0.8,\n",
    "    'bagging_freq': 5,\n",
    "    'min_child_samples': 326,\n",
    "    'min_child_weight': 10000.0,\n",
    "    'reg_alpha': 50,\n",
    "    'reg_lambda': 0,\n",
    "    'verbose': 0\n",
    "}\n",
    "\n",
    "gbm_auc = lgb.train(params,\n",
    "                lgb_train,\n",
    "                num_boost_round=20,\n",
    "                valid_sets=lgb_eval)\n",
    "                #early_stopping_rounds=100)\n",
    "\n",
    "\n",
    "\n",
    "pred_auc_gbm = gbm_auc.predict(X_test_auc, num_iteration=gbm_auc.best_iteration)\n",
    "print('The rmse of prediction is:', mean_squared_error(y_test_auc, pred_auc_gbm) ** 0.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "63315.97962183445"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "gbm_auc.save_model('model_light_gbm_auc.txt')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "_ = gbm_auc.plot_importance(model)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### LightGBM para installs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_full_installs = get_df_to_train(windows, features, label_inst, features_list_class_inst)\n",
    "df_full_installs = df_full_installs[df_full_installs[\"label_inst\"] != 259200]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train_inst, X_test_inst, y_train_inst, y_test_inst = get_train_test_split(df_full_installs, \"label_inst\", 0.05)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/lightgbm/engine.py:118: UserWarning: Found `n_estimators` in params. Will use it instead of argument\n",
      "  warnings.warn(\"Found `{}` in params. Will use it instead of argument\".format(alias))\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1]\tvalid_0's l2: 5.93669e+09\tvalid_0's l1: 67501.3\n",
      "[2]\tvalid_0's l2: 5.93505e+09\tvalid_0's l1: 67491.8\n",
      "[3]\tvalid_0's l2: 5.93367e+09\tvalid_0's l1: 67483.5\n",
      "[4]\tvalid_0's l2: 5.93221e+09\tvalid_0's l1: 67474\n",
      "[5]\tvalid_0's l2: 5.93095e+09\tvalid_0's l1: 67465.6\n",
      "[6]\tvalid_0's l2: 5.92957e+09\tvalid_0's l1: 67456.2\n",
      "[7]\tvalid_0's l2: 5.92846e+09\tvalid_0's l1: 67448.9\n",
      "[8]\tvalid_0's l2: 5.92745e+09\tvalid_0's l1: 67441.5\n",
      "[9]\tvalid_0's l2: 5.92607e+09\tvalid_0's l1: 67432.2\n",
      "[10]\tvalid_0's l2: 5.92527e+09\tvalid_0's l1: 67425.7\n",
      "[11]\tvalid_0's l2: 5.92464e+09\tvalid_0's l1: 67420.6\n",
      "[12]\tvalid_0's l2: 5.92406e+09\tvalid_0's l1: 67415.8\n",
      "[13]\tvalid_0's l2: 5.92326e+09\tvalid_0's l1: 67409.8\n",
      "[14]\tvalid_0's l2: 5.92285e+09\tvalid_0's l1: 67405.9\n",
      "[15]\tvalid_0's l2: 5.92208e+09\tvalid_0's l1: 67400.2\n",
      "[16]\tvalid_0's l2: 5.92132e+09\tvalid_0's l1: 67394.5\n",
      "[17]\tvalid_0's l2: 5.92072e+09\tvalid_0's l1: 67389.4\n",
      "[18]\tvalid_0's l2: 5.92014e+09\tvalid_0's l1: 67384.2\n",
      "[19]\tvalid_0's l2: 5.9195e+09\tvalid_0's l1: 67378.8\n",
      "[20]\tvalid_0's l2: 5.91893e+09\tvalid_0's l1: 67373.9\n",
      "[21]\tvalid_0's l2: 5.91814e+09\tvalid_0's l1: 67367.9\n",
      "[22]\tvalid_0's l2: 5.91778e+09\tvalid_0's l1: 67364.8\n",
      "[23]\tvalid_0's l2: 5.91705e+09\tvalid_0's l1: 67358.4\n",
      "[24]\tvalid_0's l2: 5.91662e+09\tvalid_0's l1: 67355.1\n",
      "[25]\tvalid_0's l2: 5.91618e+09\tvalid_0's l1: 67351\n",
      "[26]\tvalid_0's l2: 5.91567e+09\tvalid_0's l1: 67347.1\n",
      "[27]\tvalid_0's l2: 5.91509e+09\tvalid_0's l1: 67342.1\n",
      "[28]\tvalid_0's l2: 5.9147e+09\tvalid_0's l1: 67338.1\n",
      "[29]\tvalid_0's l2: 5.91434e+09\tvalid_0's l1: 67334.9\n",
      "[30]\tvalid_0's l2: 5.91391e+09\tvalid_0's l1: 67330.8\n",
      "[31]\tvalid_0's l2: 5.91367e+09\tvalid_0's l1: 67327.9\n",
      "[32]\tvalid_0's l2: 5.91335e+09\tvalid_0's l1: 67325.1\n",
      "[33]\tvalid_0's l2: 5.91306e+09\tvalid_0's l1: 67322.6\n",
      "[34]\tvalid_0's l2: 5.91277e+09\tvalid_0's l1: 67321.3\n",
      "[35]\tvalid_0's l2: 5.91264e+09\tvalid_0's l1: 67319.5\n",
      "[36]\tvalid_0's l2: 5.91231e+09\tvalid_0's l1: 67316.4\n",
      "[37]\tvalid_0's l2: 5.91212e+09\tvalid_0's l1: 67314\n",
      "[38]\tvalid_0's l2: 5.9118e+09\tvalid_0's l1: 67311\n",
      "[39]\tvalid_0's l2: 5.91137e+09\tvalid_0's l1: 67306.8\n",
      "[40]\tvalid_0's l2: 5.91126e+09\tvalid_0's l1: 67306\n",
      "[41]\tvalid_0's l2: 5.91091e+09\tvalid_0's l1: 67303.2\n",
      "[42]\tvalid_0's l2: 5.91076e+09\tvalid_0's l1: 67301.9\n",
      "[43]\tvalid_0's l2: 5.91049e+09\tvalid_0's l1: 67299.3\n",
      "[44]\tvalid_0's l2: 5.91031e+09\tvalid_0's l1: 67297.2\n",
      "[45]\tvalid_0's l2: 5.91008e+09\tvalid_0's l1: 67294.9\n",
      "[46]\tvalid_0's l2: 5.90984e+09\tvalid_0's l1: 67292.8\n",
      "[47]\tvalid_0's l2: 5.90966e+09\tvalid_0's l1: 67291.1\n",
      "[48]\tvalid_0's l2: 5.90945e+09\tvalid_0's l1: 67290\n",
      "[49]\tvalid_0's l2: 5.90916e+09\tvalid_0's l1: 67287.6\n",
      "[50]\tvalid_0's l2: 5.90911e+09\tvalid_0's l1: 67287.4\n",
      "[51]\tvalid_0's l2: 5.90903e+09\tvalid_0's l1: 67286.1\n",
      "[52]\tvalid_0's l2: 5.90899e+09\tvalid_0's l1: 67285.6\n",
      "[53]\tvalid_0's l2: 5.90899e+09\tvalid_0's l1: 67285.7\n",
      "[54]\tvalid_0's l2: 5.90893e+09\tvalid_0's l1: 67284.9\n",
      "[55]\tvalid_0's l2: 5.90881e+09\tvalid_0's l1: 67283.8\n",
      "[56]\tvalid_0's l2: 5.90872e+09\tvalid_0's l1: 67282.6\n",
      "[57]\tvalid_0's l2: 5.90863e+09\tvalid_0's l1: 67281.7\n",
      "[58]\tvalid_0's l2: 5.90849e+09\tvalid_0's l1: 67279.7\n",
      "[59]\tvalid_0's l2: 5.90837e+09\tvalid_0's l1: 67278.1\n",
      "[60]\tvalid_0's l2: 5.90832e+09\tvalid_0's l1: 67277.1\n",
      "[61]\tvalid_0's l2: 5.90841e+09\tvalid_0's l1: 67276.6\n",
      "[62]\tvalid_0's l2: 5.90835e+09\tvalid_0's l1: 67276\n",
      "[63]\tvalid_0's l2: 5.90831e+09\tvalid_0's l1: 67274.9\n",
      "[64]\tvalid_0's l2: 5.90828e+09\tvalid_0's l1: 67273.8\n",
      "[65]\tvalid_0's l2: 5.9083e+09\tvalid_0's l1: 67273.2\n",
      "[66]\tvalid_0's l2: 5.90814e+09\tvalid_0's l1: 67272.2\n",
      "[67]\tvalid_0's l2: 5.90818e+09\tvalid_0's l1: 67271.9\n",
      "[68]\tvalid_0's l2: 5.90798e+09\tvalid_0's l1: 67270.6\n",
      "[69]\tvalid_0's l2: 5.90787e+09\tvalid_0's l1: 67270.4\n",
      "[70]\tvalid_0's l2: 5.90773e+09\tvalid_0's l1: 67268.8\n",
      "[71]\tvalid_0's l2: 5.9074e+09\tvalid_0's l1: 67266.2\n",
      "[72]\tvalid_0's l2: 5.90723e+09\tvalid_0's l1: 67264.3\n",
      "[73]\tvalid_0's l2: 5.90711e+09\tvalid_0's l1: 67262.8\n",
      "[74]\tvalid_0's l2: 5.90697e+09\tvalid_0's l1: 67260.9\n",
      "[75]\tvalid_0's l2: 5.90686e+09\tvalid_0's l1: 67259.5\n",
      "[76]\tvalid_0's l2: 5.90692e+09\tvalid_0's l1: 67258.8\n",
      "[77]\tvalid_0's l2: 5.90706e+09\tvalid_0's l1: 67258.6\n",
      "[78]\tvalid_0's l2: 5.90702e+09\tvalid_0's l1: 67257.5\n",
      "[79]\tvalid_0's l2: 5.9071e+09\tvalid_0's l1: 67257.2\n",
      "[80]\tvalid_0's l2: 5.90687e+09\tvalid_0's l1: 67255.9\n",
      "[81]\tvalid_0's l2: 5.90668e+09\tvalid_0's l1: 67254.4\n",
      "[82]\tvalid_0's l2: 5.90651e+09\tvalid_0's l1: 67252.8\n",
      "[83]\tvalid_0's l2: 5.90657e+09\tvalid_0's l1: 67252.4\n",
      "[84]\tvalid_0's l2: 5.90649e+09\tvalid_0's l1: 67251.4\n",
      "[85]\tvalid_0's l2: 5.90629e+09\tvalid_0's l1: 67249.9\n",
      "[86]\tvalid_0's l2: 5.90628e+09\tvalid_0's l1: 67249.1\n",
      "[87]\tvalid_0's l2: 5.90627e+09\tvalid_0's l1: 67247.9\n",
      "[88]\tvalid_0's l2: 5.90599e+09\tvalid_0's l1: 67245.1\n",
      "[89]\tvalid_0's l2: 5.9058e+09\tvalid_0's l1: 67243\n",
      "[90]\tvalid_0's l2: 5.90574e+09\tvalid_0's l1: 67241.9\n",
      "[91]\tvalid_0's l2: 5.90566e+09\tvalid_0's l1: 67241.2\n",
      "[92]\tvalid_0's l2: 5.90568e+09\tvalid_0's l1: 67241.1\n",
      "[93]\tvalid_0's l2: 5.90574e+09\tvalid_0's l1: 67241.2\n",
      "[94]\tvalid_0's l2: 5.90584e+09\tvalid_0's l1: 67241.6\n",
      "[95]\tvalid_0's l2: 5.90573e+09\tvalid_0's l1: 67241\n",
      "[96]\tvalid_0's l2: 5.90573e+09\tvalid_0's l1: 67240.3\n",
      "[97]\tvalid_0's l2: 5.90584e+09\tvalid_0's l1: 67240.1\n",
      "[98]\tvalid_0's l2: 5.90584e+09\tvalid_0's l1: 67239.7\n",
      "[99]\tvalid_0's l2: 5.906e+09\tvalid_0's l1: 67239.9\n",
      "[100]\tvalid_0's l2: 5.90612e+09\tvalid_0's l1: 67239.8\n",
      "[101]\tvalid_0's l2: 5.90623e+09\tvalid_0's l1: 67240.7\n",
      "[102]\tvalid_0's l2: 5.90624e+09\tvalid_0's l1: 67240.5\n",
      "[103]\tvalid_0's l2: 5.90637e+09\tvalid_0's l1: 67241.2\n",
      "[104]\tvalid_0's l2: 5.90626e+09\tvalid_0's l1: 67240.6\n",
      "[105]\tvalid_0's l2: 5.90633e+09\tvalid_0's l1: 67240.8\n",
      "[106]\tvalid_0's l2: 5.90628e+09\tvalid_0's l1: 67240.3\n",
      "[107]\tvalid_0's l2: 5.90624e+09\tvalid_0's l1: 67240\n",
      "[108]\tvalid_0's l2: 5.90608e+09\tvalid_0's l1: 67239\n",
      "[109]\tvalid_0's l2: 5.90594e+09\tvalid_0's l1: 67238.1\n",
      "[110]\tvalid_0's l2: 5.90594e+09\tvalid_0's l1: 67237.9\n",
      "[111]\tvalid_0's l2: 5.90588e+09\tvalid_0's l1: 67236.8\n",
      "[112]\tvalid_0's l2: 5.90567e+09\tvalid_0's l1: 67234.3\n",
      "[113]\tvalid_0's l2: 5.90563e+09\tvalid_0's l1: 67233.7\n",
      "[114]\tvalid_0's l2: 5.90538e+09\tvalid_0's l1: 67231.8\n",
      "[115]\tvalid_0's l2: 5.90539e+09\tvalid_0's l1: 67231.2\n",
      "[116]\tvalid_0's l2: 5.90547e+09\tvalid_0's l1: 67231.9\n",
      "[117]\tvalid_0's l2: 5.90556e+09\tvalid_0's l1: 67232.8\n",
      "[118]\tvalid_0's l2: 5.90557e+09\tvalid_0's l1: 67232.9\n",
      "[119]\tvalid_0's l2: 5.90559e+09\tvalid_0's l1: 67233.4\n",
      "[120]\tvalid_0's l2: 5.90567e+09\tvalid_0's l1: 67233.6\n",
      "[121]\tvalid_0's l2: 5.90569e+09\tvalid_0's l1: 67233.7\n",
      "[122]\tvalid_0's l2: 5.90569e+09\tvalid_0's l1: 67233.6\n",
      "[123]\tvalid_0's l2: 5.90577e+09\tvalid_0's l1: 67233.9\n",
      "[124]\tvalid_0's l2: 5.90571e+09\tvalid_0's l1: 67233.6\n",
      "[125]\tvalid_0's l2: 5.90569e+09\tvalid_0's l1: 67233\n",
      "[126]\tvalid_0's l2: 5.90568e+09\tvalid_0's l1: 67232.6\n",
      "[127]\tvalid_0's l2: 5.90562e+09\tvalid_0's l1: 67232.1\n",
      "[128]\tvalid_0's l2: 5.90565e+09\tvalid_0's l1: 67232\n",
      "[129]\tvalid_0's l2: 5.90568e+09\tvalid_0's l1: 67231.9\n",
      "[130]\tvalid_0's l2: 5.90572e+09\tvalid_0's l1: 67231.7\n",
      "[131]\tvalid_0's l2: 5.90582e+09\tvalid_0's l1: 67232.2\n",
      "[132]\tvalid_0's l2: 5.90592e+09\tvalid_0's l1: 67232.3\n",
      "[133]\tvalid_0's l2: 5.90601e+09\tvalid_0's l1: 67232.4\n",
      "[134]\tvalid_0's l2: 5.90586e+09\tvalid_0's l1: 67231.4\n",
      "[135]\tvalid_0's l2: 5.90606e+09\tvalid_0's l1: 67232.6\n",
      "[136]\tvalid_0's l2: 5.90598e+09\tvalid_0's l1: 67232.1\n",
      "[137]\tvalid_0's l2: 5.90604e+09\tvalid_0's l1: 67231.8\n",
      "[138]\tvalid_0's l2: 5.90603e+09\tvalid_0's l1: 67231.7\n",
      "[139]\tvalid_0's l2: 5.90616e+09\tvalid_0's l1: 67232.1\n",
      "[140]\tvalid_0's l2: 5.90614e+09\tvalid_0's l1: 67232.2\n",
      "[141]\tvalid_0's l2: 5.90599e+09\tvalid_0's l1: 67231.3\n",
      "[142]\tvalid_0's l2: 5.90582e+09\tvalid_0's l1: 67230\n",
      "[143]\tvalid_0's l2: 5.90559e+09\tvalid_0's l1: 67228.4\n",
      "[144]\tvalid_0's l2: 5.90557e+09\tvalid_0's l1: 67228.1\n",
      "[145]\tvalid_0's l2: 5.90547e+09\tvalid_0's l1: 67227.3\n",
      "[146]\tvalid_0's l2: 5.90524e+09\tvalid_0's l1: 67225.5\n",
      "[147]\tvalid_0's l2: 5.90521e+09\tvalid_0's l1: 67224.7\n",
      "[148]\tvalid_0's l2: 5.90509e+09\tvalid_0's l1: 67223.3\n",
      "[149]\tvalid_0's l2: 5.90491e+09\tvalid_0's l1: 67221.5\n",
      "[150]\tvalid_0's l2: 5.90479e+09\tvalid_0's l1: 67220.5\n",
      "[151]\tvalid_0's l2: 5.90484e+09\tvalid_0's l1: 67221.1\n",
      "[152]\tvalid_0's l2: 5.9047e+09\tvalid_0's l1: 67220.2\n",
      "[153]\tvalid_0's l2: 5.9046e+09\tvalid_0's l1: 67219.4\n",
      "[154]\tvalid_0's l2: 5.9045e+09\tvalid_0's l1: 67218.6\n",
      "[155]\tvalid_0's l2: 5.90444e+09\tvalid_0's l1: 67218.2\n",
      "[156]\tvalid_0's l2: 5.90432e+09\tvalid_0's l1: 67217.5\n",
      "[157]\tvalid_0's l2: 5.9042e+09\tvalid_0's l1: 67215.8\n",
      "[158]\tvalid_0's l2: 5.90411e+09\tvalid_0's l1: 67215.1\n",
      "[159]\tvalid_0's l2: 5.90394e+09\tvalid_0's l1: 67213.5\n",
      "[160]\tvalid_0's l2: 5.904e+09\tvalid_0's l1: 67213.6\n",
      "[161]\tvalid_0's l2: 5.90392e+09\tvalid_0's l1: 67213\n",
      "[162]\tvalid_0's l2: 5.90379e+09\tvalid_0's l1: 67211.5\n",
      "[163]\tvalid_0's l2: 5.9039e+09\tvalid_0's l1: 67211.4\n",
      "[164]\tvalid_0's l2: 5.90378e+09\tvalid_0's l1: 67210\n",
      "[165]\tvalid_0's l2: 5.90377e+09\tvalid_0's l1: 67210\n",
      "[166]\tvalid_0's l2: 5.90375e+09\tvalid_0's l1: 67208.5\n",
      "[167]\tvalid_0's l2: 5.90377e+09\tvalid_0's l1: 67207.8\n",
      "[168]\tvalid_0's l2: 5.9038e+09\tvalid_0's l1: 67207.4\n",
      "[169]\tvalid_0's l2: 5.90384e+09\tvalid_0's l1: 67207.2\n",
      "[170]\tvalid_0's l2: 5.90392e+09\tvalid_0's l1: 67207.6\n",
      "[171]\tvalid_0's l2: 5.90385e+09\tvalid_0's l1: 67206.5\n",
      "[172]\tvalid_0's l2: 5.90386e+09\tvalid_0's l1: 67206.2\n",
      "[173]\tvalid_0's l2: 5.90385e+09\tvalid_0's l1: 67205.6\n",
      "[174]\tvalid_0's l2: 5.90378e+09\tvalid_0's l1: 67204.7\n",
      "[175]\tvalid_0's l2: 5.90368e+09\tvalid_0's l1: 67204.2\n",
      "[176]\tvalid_0's l2: 5.90381e+09\tvalid_0's l1: 67203.9\n",
      "[177]\tvalid_0's l2: 5.90385e+09\tvalid_0's l1: 67203.5\n",
      "[178]\tvalid_0's l2: 5.90393e+09\tvalid_0's l1: 67203.6\n",
      "[179]\tvalid_0's l2: 5.904e+09\tvalid_0's l1: 67203.8\n",
      "[180]\tvalid_0's l2: 5.904e+09\tvalid_0's l1: 67203.6\n",
      "[181]\tvalid_0's l2: 5.90409e+09\tvalid_0's l1: 67203.6\n",
      "[182]\tvalid_0's l2: 5.90414e+09\tvalid_0's l1: 67203.3\n",
      "[183]\tvalid_0's l2: 5.90403e+09\tvalid_0's l1: 67202.6\n",
      "[184]\tvalid_0's l2: 5.90388e+09\tvalid_0's l1: 67201.6\n",
      "[185]\tvalid_0's l2: 5.90389e+09\tvalid_0's l1: 67201.3\n",
      "[186]\tvalid_0's l2: 5.90393e+09\tvalid_0's l1: 67201.1\n",
      "[187]\tvalid_0's l2: 5.90393e+09\tvalid_0's l1: 67200.9\n",
      "[188]\tvalid_0's l2: 5.90397e+09\tvalid_0's l1: 67200.8\n",
      "[189]\tvalid_0's l2: 5.904e+09\tvalid_0's l1: 67200.6\n",
      "[190]\tvalid_0's l2: 5.90401e+09\tvalid_0's l1: 67200.4\n",
      "[191]\tvalid_0's l2: 5.90389e+09\tvalid_0's l1: 67199.9\n",
      "[192]\tvalid_0's l2: 5.90395e+09\tvalid_0's l1: 67200.2\n",
      "[193]\tvalid_0's l2: 5.90374e+09\tvalid_0's l1: 67198.9\n",
      "[194]\tvalid_0's l2: 5.90371e+09\tvalid_0's l1: 67198.6\n",
      "[195]\tvalid_0's l2: 5.90366e+09\tvalid_0's l1: 67198.4\n",
      "[196]\tvalid_0's l2: 5.90363e+09\tvalid_0's l1: 67197.9\n",
      "[197]\tvalid_0's l2: 5.90373e+09\tvalid_0's l1: 67198.6\n",
      "[198]\tvalid_0's l2: 5.90368e+09\tvalid_0's l1: 67198.9\n",
      "[199]\tvalid_0's l2: 5.9035e+09\tvalid_0's l1: 67197.5\n",
      "[200]\tvalid_0's l2: 5.9035e+09\tvalid_0's l1: 67197.4\n",
      "[201]\tvalid_0's l2: 5.90345e+09\tvalid_0's l1: 67197.3\n",
      "[202]\tvalid_0's l2: 5.9034e+09\tvalid_0's l1: 67196.8\n",
      "[203]\tvalid_0's l2: 5.90337e+09\tvalid_0's l1: 67196.5\n",
      "[204]\tvalid_0's l2: 5.90333e+09\tvalid_0's l1: 67196.3\n",
      "[205]\tvalid_0's l2: 5.90327e+09\tvalid_0's l1: 67195.9\n",
      "[206]\tvalid_0's l2: 5.90328e+09\tvalid_0's l1: 67195.4\n",
      "[207]\tvalid_0's l2: 5.90327e+09\tvalid_0's l1: 67194.7\n",
      "[208]\tvalid_0's l2: 5.90324e+09\tvalid_0's l1: 67193.8\n",
      "[209]\tvalid_0's l2: 5.90319e+09\tvalid_0's l1: 67192.8\n",
      "[210]\tvalid_0's l2: 5.90313e+09\tvalid_0's l1: 67192.1\n",
      "[211]\tvalid_0's l2: 5.90298e+09\tvalid_0's l1: 67190.8\n",
      "[212]\tvalid_0's l2: 5.90301e+09\tvalid_0's l1: 67191\n",
      "[213]\tvalid_0's l2: 5.90307e+09\tvalid_0's l1: 67191.1\n",
      "[214]\tvalid_0's l2: 5.90302e+09\tvalid_0's l1: 67190.5\n",
      "[215]\tvalid_0's l2: 5.903e+09\tvalid_0's l1: 67190.4\n",
      "[216]\tvalid_0's l2: 5.90306e+09\tvalid_0's l1: 67190.5\n",
      "[217]\tvalid_0's l2: 5.90314e+09\tvalid_0's l1: 67190.8\n",
      "[218]\tvalid_0's l2: 5.90318e+09\tvalid_0's l1: 67191.2\n",
      "[219]\tvalid_0's l2: 5.90327e+09\tvalid_0's l1: 67191.7\n",
      "[220]\tvalid_0's l2: 5.90324e+09\tvalid_0's l1: 67191.5\n",
      "[221]\tvalid_0's l2: 5.90336e+09\tvalid_0's l1: 67191.8\n",
      "[222]\tvalid_0's l2: 5.90335e+09\tvalid_0's l1: 67191.8\n",
      "[223]\tvalid_0's l2: 5.90326e+09\tvalid_0's l1: 67191.3\n",
      "[224]\tvalid_0's l2: 5.90318e+09\tvalid_0's l1: 67190.7\n",
      "[225]\tvalid_0's l2: 5.90301e+09\tvalid_0's l1: 67189.6\n",
      "[226]\tvalid_0's l2: 5.90308e+09\tvalid_0's l1: 67189.3\n",
      "[227]\tvalid_0's l2: 5.90311e+09\tvalid_0's l1: 67189.1\n",
      "[228]\tvalid_0's l2: 5.90319e+09\tvalid_0's l1: 67189.5\n",
      "[229]\tvalid_0's l2: 5.90321e+09\tvalid_0's l1: 67188.9\n",
      "[230]\tvalid_0's l2: 5.90315e+09\tvalid_0's l1: 67188.2\n",
      "[231]\tvalid_0's l2: 5.90306e+09\tvalid_0's l1: 67187.5\n",
      "[232]\tvalid_0's l2: 5.90309e+09\tvalid_0's l1: 67186.8\n",
      "[233]\tvalid_0's l2: 5.90305e+09\tvalid_0's l1: 67186.5\n",
      "[234]\tvalid_0's l2: 5.90294e+09\tvalid_0's l1: 67185\n",
      "[235]\tvalid_0's l2: 5.90287e+09\tvalid_0's l1: 67184.3\n",
      "[236]\tvalid_0's l2: 5.90284e+09\tvalid_0's l1: 67183.8\n",
      "[237]\tvalid_0's l2: 5.90297e+09\tvalid_0's l1: 67184.5\n",
      "[238]\tvalid_0's l2: 5.90305e+09\tvalid_0's l1: 67184.8\n",
      "[239]\tvalid_0's l2: 5.90303e+09\tvalid_0's l1: 67184.3\n",
      "[240]\tvalid_0's l2: 5.90312e+09\tvalid_0's l1: 67184.6\n",
      "[241]\tvalid_0's l2: 5.90303e+09\tvalid_0's l1: 67183.9\n",
      "[242]\tvalid_0's l2: 5.90296e+09\tvalid_0's l1: 67183\n",
      "[243]\tvalid_0's l2: 5.90297e+09\tvalid_0's l1: 67183\n",
      "[244]\tvalid_0's l2: 5.90288e+09\tvalid_0's l1: 67182.2\n",
      "[245]\tvalid_0's l2: 5.90275e+09\tvalid_0's l1: 67181.2\n",
      "[246]\tvalid_0's l2: 5.90277e+09\tvalid_0's l1: 67181.5\n",
      "[247]\tvalid_0's l2: 5.90276e+09\tvalid_0's l1: 67181.3\n",
      "[248]\tvalid_0's l2: 5.90282e+09\tvalid_0's l1: 67181.6\n",
      "[249]\tvalid_0's l2: 5.90282e+09\tvalid_0's l1: 67181.4\n",
      "[250]\tvalid_0's l2: 5.90289e+09\tvalid_0's l1: 67181.7\n",
      "[251]\tvalid_0's l2: 5.90281e+09\tvalid_0's l1: 67181\n",
      "[252]\tvalid_0's l2: 5.90278e+09\tvalid_0's l1: 67180.7\n",
      "[253]\tvalid_0's l2: 5.90275e+09\tvalid_0's l1: 67180.2\n",
      "[254]\tvalid_0's l2: 5.9025e+09\tvalid_0's l1: 67179.1\n",
      "[255]\tvalid_0's l2: 5.90251e+09\tvalid_0's l1: 67178.9\n",
      "[256]\tvalid_0's l2: 5.90263e+09\tvalid_0's l1: 67179.4\n",
      "[257]\tvalid_0's l2: 5.90253e+09\tvalid_0's l1: 67178.6\n",
      "[258]\tvalid_0's l2: 5.90264e+09\tvalid_0's l1: 67178.8\n",
      "[259]\tvalid_0's l2: 5.90261e+09\tvalid_0's l1: 67178.3\n",
      "[260]\tvalid_0's l2: 5.90272e+09\tvalid_0's l1: 67178.5\n",
      "[261]\tvalid_0's l2: 5.90258e+09\tvalid_0's l1: 67177.8\n",
      "[262]\tvalid_0's l2: 5.90264e+09\tvalid_0's l1: 67178.6\n",
      "[263]\tvalid_0's l2: 5.90285e+09\tvalid_0's l1: 67179.9\n",
      "[264]\tvalid_0's l2: 5.90269e+09\tvalid_0's l1: 67178.9\n",
      "[265]\tvalid_0's l2: 5.90268e+09\tvalid_0's l1: 67179.3\n",
      "[266]\tvalid_0's l2: 5.90265e+09\tvalid_0's l1: 67179.3\n",
      "[267]\tvalid_0's l2: 5.90255e+09\tvalid_0's l1: 67178.3\n",
      "[268]\tvalid_0's l2: 5.90251e+09\tvalid_0's l1: 67177.8\n",
      "[269]\tvalid_0's l2: 5.90244e+09\tvalid_0's l1: 67177.1\n",
      "[270]\tvalid_0's l2: 5.90248e+09\tvalid_0's l1: 67177\n",
      "[271]\tvalid_0's l2: 5.90234e+09\tvalid_0's l1: 67175.6\n",
      "[272]\tvalid_0's l2: 5.90212e+09\tvalid_0's l1: 67174.2\n",
      "[273]\tvalid_0's l2: 5.90208e+09\tvalid_0's l1: 67173.3\n",
      "[274]\tvalid_0's l2: 5.90215e+09\tvalid_0's l1: 67173.7\n",
      "[275]\tvalid_0's l2: 5.9021e+09\tvalid_0's l1: 67173.4\n",
      "[276]\tvalid_0's l2: 5.90208e+09\tvalid_0's l1: 67173.1\n",
      "[277]\tvalid_0's l2: 5.90204e+09\tvalid_0's l1: 67172.8\n",
      "[278]\tvalid_0's l2: 5.90201e+09\tvalid_0's l1: 67172.4\n",
      "[279]\tvalid_0's l2: 5.90201e+09\tvalid_0's l1: 67172.4\n",
      "[280]\tvalid_0's l2: 5.90197e+09\tvalid_0's l1: 67172.1\n",
      "[281]\tvalid_0's l2: 5.90187e+09\tvalid_0's l1: 67171.4\n",
      "[282]\tvalid_0's l2: 5.90186e+09\tvalid_0's l1: 67171.3\n",
      "[283]\tvalid_0's l2: 5.90191e+09\tvalid_0's l1: 67171.8\n",
      "[284]\tvalid_0's l2: 5.90172e+09\tvalid_0's l1: 67170.7\n",
      "[285]\tvalid_0's l2: 5.90173e+09\tvalid_0's l1: 67170.5\n",
      "[286]\tvalid_0's l2: 5.90171e+09\tvalid_0's l1: 67170.5\n",
      "[287]\tvalid_0's l2: 5.90163e+09\tvalid_0's l1: 67169.7\n",
      "[288]\tvalid_0's l2: 5.90161e+09\tvalid_0's l1: 67169.6\n",
      "[289]\tvalid_0's l2: 5.90153e+09\tvalid_0's l1: 67169\n",
      "[290]\tvalid_0's l2: 5.90146e+09\tvalid_0's l1: 67168.3\n",
      "[291]\tvalid_0's l2: 5.90124e+09\tvalid_0's l1: 67166.8\n",
      "[292]\tvalid_0's l2: 5.9012e+09\tvalid_0's l1: 67166.5\n",
      "[293]\tvalid_0's l2: 5.90106e+09\tvalid_0's l1: 67165.3\n",
      "[294]\tvalid_0's l2: 5.90106e+09\tvalid_0's l1: 67165.1\n",
      "[295]\tvalid_0's l2: 5.90113e+09\tvalid_0's l1: 67165.6\n",
      "[296]\tvalid_0's l2: 5.90111e+09\tvalid_0's l1: 67165.7\n",
      "[297]\tvalid_0's l2: 5.90102e+09\tvalid_0's l1: 67165.4\n",
      "[298]\tvalid_0's l2: 5.90105e+09\tvalid_0's l1: 67165.9\n",
      "[299]\tvalid_0's l2: 5.90107e+09\tvalid_0's l1: 67166.6\n",
      "[300]\tvalid_0's l2: 5.90099e+09\tvalid_0's l1: 67165.9\n",
      "[301]\tvalid_0's l2: 5.90101e+09\tvalid_0's l1: 67166\n",
      "[302]\tvalid_0's l2: 5.90115e+09\tvalid_0's l1: 67167\n",
      "[303]\tvalid_0's l2: 5.90115e+09\tvalid_0's l1: 67166.8\n",
      "[304]\tvalid_0's l2: 5.90129e+09\tvalid_0's l1: 67167.5\n",
      "[305]\tvalid_0's l2: 5.90137e+09\tvalid_0's l1: 67167.6\n",
      "[306]\tvalid_0's l2: 5.90138e+09\tvalid_0's l1: 67166.8\n",
      "[307]\tvalid_0's l2: 5.90135e+09\tvalid_0's l1: 67166.4\n",
      "[308]\tvalid_0's l2: 5.90129e+09\tvalid_0's l1: 67165.6\n",
      "[309]\tvalid_0's l2: 5.90133e+09\tvalid_0's l1: 67165.5\n",
      "[310]\tvalid_0's l2: 5.90135e+09\tvalid_0's l1: 67165.3\n",
      "[311]\tvalid_0's l2: 5.90136e+09\tvalid_0's l1: 67165.1\n",
      "[312]\tvalid_0's l2: 5.90136e+09\tvalid_0's l1: 67165.2\n",
      "[313]\tvalid_0's l2: 5.9014e+09\tvalid_0's l1: 67165.5\n",
      "[314]\tvalid_0's l2: 5.90147e+09\tvalid_0's l1: 67166.3\n",
      "[315]\tvalid_0's l2: 5.9015e+09\tvalid_0's l1: 67166.7\n",
      "[316]\tvalid_0's l2: 5.90146e+09\tvalid_0's l1: 67166.2\n",
      "[317]\tvalid_0's l2: 5.90149e+09\tvalid_0's l1: 67166.6\n",
      "[318]\tvalid_0's l2: 5.90157e+09\tvalid_0's l1: 67167.2\n",
      "[319]\tvalid_0's l2: 5.90156e+09\tvalid_0's l1: 67166.8\n",
      "[320]\tvalid_0's l2: 5.90161e+09\tvalid_0's l1: 67166.9\n",
      "[321]\tvalid_0's l2: 5.90157e+09\tvalid_0's l1: 67166.2\n",
      "[322]\tvalid_0's l2: 5.90159e+09\tvalid_0's l1: 67165.7\n",
      "[323]\tvalid_0's l2: 5.90164e+09\tvalid_0's l1: 67165.8\n",
      "[324]\tvalid_0's l2: 5.90171e+09\tvalid_0's l1: 67166.2\n",
      "[325]\tvalid_0's l2: 5.90172e+09\tvalid_0's l1: 67166.1\n",
      "[326]\tvalid_0's l2: 5.9018e+09\tvalid_0's l1: 67166.1\n",
      "[327]\tvalid_0's l2: 5.90183e+09\tvalid_0's l1: 67166.3\n",
      "[328]\tvalid_0's l2: 5.9018e+09\tvalid_0's l1: 67165.6\n",
      "[329]\tvalid_0's l2: 5.90181e+09\tvalid_0's l1: 67165.2\n",
      "[330]\tvalid_0's l2: 5.90181e+09\tvalid_0's l1: 67164.6\n",
      "[331]\tvalid_0's l2: 5.90168e+09\tvalid_0's l1: 67163.9\n",
      "[332]\tvalid_0's l2: 5.90152e+09\tvalid_0's l1: 67163.2\n",
      "[333]\tvalid_0's l2: 5.90162e+09\tvalid_0's l1: 67163.8\n",
      "[334]\tvalid_0's l2: 5.90174e+09\tvalid_0's l1: 67164.4\n",
      "[335]\tvalid_0's l2: 5.90186e+09\tvalid_0's l1: 67164.9\n",
      "[336]\tvalid_0's l2: 5.90188e+09\tvalid_0's l1: 67164.7\n",
      "[337]\tvalid_0's l2: 5.90173e+09\tvalid_0's l1: 67163.4\n",
      "[338]\tvalid_0's l2: 5.90162e+09\tvalid_0's l1: 67162.3\n",
      "[339]\tvalid_0's l2: 5.90163e+09\tvalid_0's l1: 67162.2\n",
      "[340]\tvalid_0's l2: 5.90161e+09\tvalid_0's l1: 67161.2\n",
      "[341]\tvalid_0's l2: 5.90149e+09\tvalid_0's l1: 67160.2\n",
      "[342]\tvalid_0's l2: 5.90157e+09\tvalid_0's l1: 67160.6\n",
      "[343]\tvalid_0's l2: 5.90139e+09\tvalid_0's l1: 67159.1\n",
      "[344]\tvalid_0's l2: 5.9014e+09\tvalid_0's l1: 67159.1\n",
      "[345]\tvalid_0's l2: 5.90129e+09\tvalid_0's l1: 67158.4\n",
      "[346]\tvalid_0's l2: 5.90131e+09\tvalid_0's l1: 67158.1\n",
      "[347]\tvalid_0's l2: 5.90133e+09\tvalid_0's l1: 67157.8\n",
      "[348]\tvalid_0's l2: 5.90137e+09\tvalid_0's l1: 67157.8\n",
      "[349]\tvalid_0's l2: 5.90138e+09\tvalid_0's l1: 67157.9\n",
      "[350]\tvalid_0's l2: 5.90137e+09\tvalid_0's l1: 67157.4\n",
      "[351]\tvalid_0's l2: 5.90138e+09\tvalid_0's l1: 67157.1\n",
      "[352]\tvalid_0's l2: 5.90127e+09\tvalid_0's l1: 67156\n",
      "[353]\tvalid_0's l2: 5.90125e+09\tvalid_0's l1: 67155.7\n",
      "[354]\tvalid_0's l2: 5.9012e+09\tvalid_0's l1: 67154.7\n",
      "[355]\tvalid_0's l2: 5.90103e+09\tvalid_0's l1: 67153.6\n",
      "[356]\tvalid_0's l2: 5.90105e+09\tvalid_0's l1: 67153.2\n",
      "[357]\tvalid_0's l2: 5.90102e+09\tvalid_0's l1: 67152.5\n",
      "[358]\tvalid_0's l2: 5.90106e+09\tvalid_0's l1: 67152.1\n",
      "[359]\tvalid_0's l2: 5.90099e+09\tvalid_0's l1: 67151\n",
      "[360]\tvalid_0's l2: 5.90094e+09\tvalid_0's l1: 67150.5\n",
      "[361]\tvalid_0's l2: 5.90084e+09\tvalid_0's l1: 67149.5\n",
      "[362]\tvalid_0's l2: 5.90085e+09\tvalid_0's l1: 67149.1\n",
      "[363]\tvalid_0's l2: 5.90093e+09\tvalid_0's l1: 67149.4\n",
      "[364]\tvalid_0's l2: 5.90092e+09\tvalid_0's l1: 67149.5\n",
      "[365]\tvalid_0's l2: 5.90094e+09\tvalid_0's l1: 67149.7\n",
      "[366]\tvalid_0's l2: 5.90093e+09\tvalid_0's l1: 67149.4\n",
      "[367]\tvalid_0's l2: 5.90102e+09\tvalid_0's l1: 67149.9\n",
      "[368]\tvalid_0's l2: 5.90091e+09\tvalid_0's l1: 67149\n",
      "[369]\tvalid_0's l2: 5.90095e+09\tvalid_0's l1: 67148.7\n",
      "[370]\tvalid_0's l2: 5.90109e+09\tvalid_0's l1: 67149.6\n",
      "[371]\tvalid_0's l2: 5.90101e+09\tvalid_0's l1: 67148.6\n",
      "[372]\tvalid_0's l2: 5.90094e+09\tvalid_0's l1: 67147.7\n",
      "[373]\tvalid_0's l2: 5.90095e+09\tvalid_0's l1: 67147.5\n",
      "[374]\tvalid_0's l2: 5.90086e+09\tvalid_0's l1: 67146.7\n",
      "[375]\tvalid_0's l2: 5.90072e+09\tvalid_0's l1: 67145.7\n",
      "[376]\tvalid_0's l2: 5.90083e+09\tvalid_0's l1: 67146\n",
      "[377]\tvalid_0's l2: 5.90088e+09\tvalid_0's l1: 67146.1\n",
      "[378]\tvalid_0's l2: 5.9009e+09\tvalid_0's l1: 67146.1\n",
      "[379]\tvalid_0's l2: 5.90097e+09\tvalid_0's l1: 67146.6\n",
      "[380]\tvalid_0's l2: 5.90109e+09\tvalid_0's l1: 67147.4\n",
      "[381]\tvalid_0's l2: 5.90117e+09\tvalid_0's l1: 67148\n",
      "[382]\tvalid_0's l2: 5.90103e+09\tvalid_0's l1: 67147\n",
      "[383]\tvalid_0's l2: 5.90108e+09\tvalid_0's l1: 67147.6\n",
      "[384]\tvalid_0's l2: 5.90098e+09\tvalid_0's l1: 67147\n",
      "[385]\tvalid_0's l2: 5.90092e+09\tvalid_0's l1: 67146.3\n",
      "[386]\tvalid_0's l2: 5.90081e+09\tvalid_0's l1: 67145.2\n",
      "[387]\tvalid_0's l2: 5.90072e+09\tvalid_0's l1: 67144.4\n",
      "[388]\tvalid_0's l2: 5.90081e+09\tvalid_0's l1: 67145.4\n",
      "[389]\tvalid_0's l2: 5.90078e+09\tvalid_0's l1: 67145.3\n",
      "[390]\tvalid_0's l2: 5.90087e+09\tvalid_0's l1: 67145.8\n",
      "[391]\tvalid_0's l2: 5.90092e+09\tvalid_0's l1: 67145.7\n",
      "[392]\tvalid_0's l2: 5.90092e+09\tvalid_0's l1: 67145.6\n",
      "[393]\tvalid_0's l2: 5.90086e+09\tvalid_0's l1: 67145.3\n",
      "[394]\tvalid_0's l2: 5.90093e+09\tvalid_0's l1: 67145.4\n",
      "[395]\tvalid_0's l2: 5.90099e+09\tvalid_0's l1: 67145.7\n",
      "[396]\tvalid_0's l2: 5.90105e+09\tvalid_0's l1: 67146.3\n",
      "[397]\tvalid_0's l2: 5.90104e+09\tvalid_0's l1: 67146.4\n",
      "[398]\tvalid_0's l2: 5.90114e+09\tvalid_0's l1: 67147\n",
      "[399]\tvalid_0's l2: 5.90126e+09\tvalid_0's l1: 67147.5\n",
      "[400]\tvalid_0's l2: 5.9012e+09\tvalid_0's l1: 67147.3\n",
      "[401]\tvalid_0's l2: 5.90124e+09\tvalid_0's l1: 67147.8\n",
      "[402]\tvalid_0's l2: 5.90127e+09\tvalid_0's l1: 67148.2\n",
      "[403]\tvalid_0's l2: 5.90127e+09\tvalid_0's l1: 67148.1\n",
      "[404]\tvalid_0's l2: 5.90134e+09\tvalid_0's l1: 67148.6\n",
      "[405]\tvalid_0's l2: 5.9014e+09\tvalid_0's l1: 67149.2\n",
      "[406]\tvalid_0's l2: 5.90143e+09\tvalid_0's l1: 67149\n",
      "[407]\tvalid_0's l2: 5.90147e+09\tvalid_0's l1: 67148.9\n",
      "[408]\tvalid_0's l2: 5.90139e+09\tvalid_0's l1: 67148.3\n",
      "[409]\tvalid_0's l2: 5.90132e+09\tvalid_0's l1: 67147.6\n",
      "[410]\tvalid_0's l2: 5.90127e+09\tvalid_0's l1: 67147\n",
      "[411]\tvalid_0's l2: 5.9013e+09\tvalid_0's l1: 67147\n",
      "[412]\tvalid_0's l2: 5.9013e+09\tvalid_0's l1: 67146.9\n",
      "[413]\tvalid_0's l2: 5.90136e+09\tvalid_0's l1: 67147.3\n",
      "[414]\tvalid_0's l2: 5.9014e+09\tvalid_0's l1: 67147.7\n",
      "[415]\tvalid_0's l2: 5.90133e+09\tvalid_0's l1: 67147.5\n",
      "[416]\tvalid_0's l2: 5.90139e+09\tvalid_0's l1: 67147.8\n",
      "[417]\tvalid_0's l2: 5.90142e+09\tvalid_0's l1: 67147.5\n",
      "[418]\tvalid_0's l2: 5.9015e+09\tvalid_0's l1: 67147.8\n",
      "[419]\tvalid_0's l2: 5.90145e+09\tvalid_0's l1: 67147.3\n",
      "[420]\tvalid_0's l2: 5.90141e+09\tvalid_0's l1: 67146.3\n",
      "[421]\tvalid_0's l2: 5.90152e+09\tvalid_0's l1: 67147.2\n",
      "[422]\tvalid_0's l2: 5.90135e+09\tvalid_0's l1: 67147\n",
      "[423]\tvalid_0's l2: 5.90143e+09\tvalid_0's l1: 67147.8\n",
      "[424]\tvalid_0's l2: 5.90138e+09\tvalid_0's l1: 67147.6\n",
      "[425]\tvalid_0's l2: 5.90136e+09\tvalid_0's l1: 67147.6\n",
      "[426]\tvalid_0's l2: 5.90139e+09\tvalid_0's l1: 67147.7\n",
      "[427]\tvalid_0's l2: 5.90132e+09\tvalid_0's l1: 67147.3\n",
      "[428]\tvalid_0's l2: 5.90137e+09\tvalid_0's l1: 67147.6\n",
      "[429]\tvalid_0's l2: 5.90128e+09\tvalid_0's l1: 67147.2\n",
      "[430]\tvalid_0's l2: 5.90126e+09\tvalid_0's l1: 67147.1\n",
      "[431]\tvalid_0's l2: 5.90124e+09\tvalid_0's l1: 67146.4\n",
      "[432]\tvalid_0's l2: 5.90124e+09\tvalid_0's l1: 67145.9\n",
      "[433]\tvalid_0's l2: 5.90127e+09\tvalid_0's l1: 67145.2\n",
      "[434]\tvalid_0's l2: 5.90133e+09\tvalid_0's l1: 67145.1\n",
      "[435]\tvalid_0's l2: 5.90125e+09\tvalid_0's l1: 67144.5\n",
      "[436]\tvalid_0's l2: 5.90138e+09\tvalid_0's l1: 67144.8\n",
      "[437]\tvalid_0's l2: 5.90145e+09\tvalid_0's l1: 67144.8\n",
      "[438]\tvalid_0's l2: 5.90153e+09\tvalid_0's l1: 67144.6\n",
      "[439]\tvalid_0's l2: 5.9016e+09\tvalid_0's l1: 67144.6\n",
      "[440]\tvalid_0's l2: 5.90171e+09\tvalid_0's l1: 67145.1\n",
      "[441]\tvalid_0's l2: 5.90174e+09\tvalid_0's l1: 67145.5\n",
      "[442]\tvalid_0's l2: 5.90185e+09\tvalid_0's l1: 67146.1\n",
      "[443]\tvalid_0's l2: 5.90199e+09\tvalid_0's l1: 67146.5\n",
      "[444]\tvalid_0's l2: 5.902e+09\tvalid_0's l1: 67146.7\n",
      "[445]\tvalid_0's l2: 5.90212e+09\tvalid_0's l1: 67147.2\n",
      "[446]\tvalid_0's l2: 5.90219e+09\tvalid_0's l1: 67147.2\n",
      "[447]\tvalid_0's l2: 5.90224e+09\tvalid_0's l1: 67147.5\n",
      "[448]\tvalid_0's l2: 5.9023e+09\tvalid_0's l1: 67147.8\n",
      "[449]\tvalid_0's l2: 5.90236e+09\tvalid_0's l1: 67147.8\n",
      "[450]\tvalid_0's l2: 5.90241e+09\tvalid_0's l1: 67147.9\n",
      "[451]\tvalid_0's l2: 5.90239e+09\tvalid_0's l1: 67147.6\n",
      "[452]\tvalid_0's l2: 5.90238e+09\tvalid_0's l1: 67147\n",
      "[453]\tvalid_0's l2: 5.90237e+09\tvalid_0's l1: 67146.8\n",
      "[454]\tvalid_0's l2: 5.90246e+09\tvalid_0's l1: 67146.9\n",
      "[455]\tvalid_0's l2: 5.90263e+09\tvalid_0's l1: 67147.5\n",
      "[456]\tvalid_0's l2: 5.9026e+09\tvalid_0's l1: 67147.4\n",
      "[457]\tvalid_0's l2: 5.9025e+09\tvalid_0's l1: 67146.9\n",
      "[458]\tvalid_0's l2: 5.90246e+09\tvalid_0's l1: 67146.7\n",
      "[459]\tvalid_0's l2: 5.90242e+09\tvalid_0's l1: 67146.6\n",
      "[460]\tvalid_0's l2: 5.90245e+09\tvalid_0's l1: 67146.7\n",
      "[461]\tvalid_0's l2: 5.90252e+09\tvalid_0's l1: 67146.9\n",
      "[462]\tvalid_0's l2: 5.90252e+09\tvalid_0's l1: 67146.8\n",
      "[463]\tvalid_0's l2: 5.9025e+09\tvalid_0's l1: 67146.9\n",
      "[464]\tvalid_0's l2: 5.90236e+09\tvalid_0's l1: 67146\n",
      "[465]\tvalid_0's l2: 5.9024e+09\tvalid_0's l1: 67146.6\n",
      "[466]\tvalid_0's l2: 5.90241e+09\tvalid_0's l1: 67146.8\n",
      "[467]\tvalid_0's l2: 5.90238e+09\tvalid_0's l1: 67146.8\n",
      "[468]\tvalid_0's l2: 5.90235e+09\tvalid_0's l1: 67146.8\n",
      "[469]\tvalid_0's l2: 5.90241e+09\tvalid_0's l1: 67147.4\n",
      "[470]\tvalid_0's l2: 5.90232e+09\tvalid_0's l1: 67147\n",
      "[471]\tvalid_0's l2: 5.90247e+09\tvalid_0's l1: 67147.6\n",
      "[472]\tvalid_0's l2: 5.90267e+09\tvalid_0's l1: 67148.8\n",
      "[473]\tvalid_0's l2: 5.90268e+09\tvalid_0's l1: 67148.7\n",
      "[474]\tvalid_0's l2: 5.90271e+09\tvalid_0's l1: 67148.6\n",
      "[475]\tvalid_0's l2: 5.90277e+09\tvalid_0's l1: 67148.5\n",
      "[476]\tvalid_0's l2: 5.90292e+09\tvalid_0's l1: 67149\n",
      "[477]\tvalid_0's l2: 5.90296e+09\tvalid_0's l1: 67148.8\n",
      "[478]\tvalid_0's l2: 5.90308e+09\tvalid_0's l1: 67149.1\n",
      "[479]\tvalid_0's l2: 5.90315e+09\tvalid_0's l1: 67148.9\n",
      "[480]\tvalid_0's l2: 5.90326e+09\tvalid_0's l1: 67149.4\n",
      "[481]\tvalid_0's l2: 5.90323e+09\tvalid_0's l1: 67149\n",
      "[482]\tvalid_0's l2: 5.90318e+09\tvalid_0's l1: 67148.5\n",
      "[483]\tvalid_0's l2: 5.90319e+09\tvalid_0's l1: 67148.5\n",
      "[484]\tvalid_0's l2: 5.90318e+09\tvalid_0's l1: 67148.3\n",
      "[485]\tvalid_0's l2: 5.90317e+09\tvalid_0's l1: 67148.2\n",
      "[486]\tvalid_0's l2: 5.9033e+09\tvalid_0's l1: 67149.1\n",
      "[487]\tvalid_0's l2: 5.90311e+09\tvalid_0's l1: 67148.5\n",
      "[488]\tvalid_0's l2: 5.90319e+09\tvalid_0's l1: 67149.1\n",
      "[489]\tvalid_0's l2: 5.90323e+09\tvalid_0's l1: 67149.1\n",
      "[490]\tvalid_0's l2: 5.9032e+09\tvalid_0's l1: 67148.7\n",
      "[491]\tvalid_0's l2: 5.9032e+09\tvalid_0's l1: 67149.3\n",
      "[492]\tvalid_0's l2: 5.90322e+09\tvalid_0's l1: 67149.6\n",
      "[493]\tvalid_0's l2: 5.90327e+09\tvalid_0's l1: 67149.8\n",
      "[494]\tvalid_0's l2: 5.90332e+09\tvalid_0's l1: 67150.1\n",
      "[495]\tvalid_0's l2: 5.90316e+09\tvalid_0's l1: 67149.2\n",
      "[496]\tvalid_0's l2: 5.90308e+09\tvalid_0's l1: 67148.7\n",
      "[497]\tvalid_0's l2: 5.90313e+09\tvalid_0's l1: 67148.7\n",
      "[498]\tvalid_0's l2: 5.90317e+09\tvalid_0's l1: 67148.7\n",
      "[499]\tvalid_0's l2: 5.90317e+09\tvalid_0's l1: 67148.8\n",
      "[500]\tvalid_0's l2: 5.9032e+09\tvalid_0's l1: 67149\n",
      "[501]\tvalid_0's l2: 5.90325e+09\tvalid_0's l1: 67149.6\n",
      "[502]\tvalid_0's l2: 5.90333e+09\tvalid_0's l1: 67150.2\n",
      "[503]\tvalid_0's l2: 5.90331e+09\tvalid_0's l1: 67150.2\n",
      "[504]\tvalid_0's l2: 5.9033e+09\tvalid_0's l1: 67150.1\n",
      "[505]\tvalid_0's l2: 5.90338e+09\tvalid_0's l1: 67150.4\n",
      "[506]\tvalid_0's l2: 5.90345e+09\tvalid_0's l1: 67150.6\n",
      "[507]\tvalid_0's l2: 5.90353e+09\tvalid_0's l1: 67151.6\n",
      "[508]\tvalid_0's l2: 5.90358e+09\tvalid_0's l1: 67152.2\n",
      "[509]\tvalid_0's l2: 5.90363e+09\tvalid_0's l1: 67152.8\n",
      "[510]\tvalid_0's l2: 5.90358e+09\tvalid_0's l1: 67153\n",
      "[511]\tvalid_0's l2: 5.90361e+09\tvalid_0's l1: 67153.1\n",
      "[512]\tvalid_0's l2: 5.90368e+09\tvalid_0's l1: 67153.2\n",
      "[513]\tvalid_0's l2: 5.9037e+09\tvalid_0's l1: 67152.9\n",
      "[514]\tvalid_0's l2: 5.90359e+09\tvalid_0's l1: 67152.5\n",
      "[515]\tvalid_0's l2: 5.90368e+09\tvalid_0's l1: 67152.9\n",
      "[516]\tvalid_0's l2: 5.90364e+09\tvalid_0's l1: 67151.8\n",
      "[517]\tvalid_0's l2: 5.9037e+09\tvalid_0's l1: 67151\n",
      "[518]\tvalid_0's l2: 5.90378e+09\tvalid_0's l1: 67150.6\n",
      "[519]\tvalid_0's l2: 5.90382e+09\tvalid_0's l1: 67150.4\n",
      "[520]\tvalid_0's l2: 5.9038e+09\tvalid_0's l1: 67150\n",
      "[521]\tvalid_0's l2: 5.90382e+09\tvalid_0's l1: 67150.5\n",
      "[522]\tvalid_0's l2: 5.9038e+09\tvalid_0's l1: 67150.5\n",
      "[523]\tvalid_0's l2: 5.90379e+09\tvalid_0's l1: 67150.6\n",
      "[524]\tvalid_0's l2: 5.90372e+09\tvalid_0's l1: 67150.2\n",
      "[525]\tvalid_0's l2: 5.90365e+09\tvalid_0's l1: 67149.7\n",
      "[526]\tvalid_0's l2: 5.90358e+09\tvalid_0's l1: 67149.4\n",
      "[527]\tvalid_0's l2: 5.90355e+09\tvalid_0's l1: 67149\n",
      "[528]\tvalid_0's l2: 5.9035e+09\tvalid_0's l1: 67148.7\n",
      "[529]\tvalid_0's l2: 5.90345e+09\tvalid_0's l1: 67148.6\n",
      "[530]\tvalid_0's l2: 5.90343e+09\tvalid_0's l1: 67148.5\n",
      "[531]\tvalid_0's l2: 5.90352e+09\tvalid_0's l1: 67148.9\n",
      "[532]\tvalid_0's l2: 5.90347e+09\tvalid_0's l1: 67148.5\n",
      "[533]\tvalid_0's l2: 5.90335e+09\tvalid_0's l1: 67147.6\n",
      "[534]\tvalid_0's l2: 5.90329e+09\tvalid_0's l1: 67146.8\n",
      "[535]\tvalid_0's l2: 5.90321e+09\tvalid_0's l1: 67145.7\n",
      "[536]\tvalid_0's l2: 5.90322e+09\tvalid_0's l1: 67145.4\n",
      "[537]\tvalid_0's l2: 5.90326e+09\tvalid_0's l1: 67145.4\n",
      "[538]\tvalid_0's l2: 5.90326e+09\tvalid_0's l1: 67145.1\n",
      "[539]\tvalid_0's l2: 5.90342e+09\tvalid_0's l1: 67145.4\n",
      "[540]\tvalid_0's l2: 5.90332e+09\tvalid_0's l1: 67144.8\n",
      "[541]\tvalid_0's l2: 5.90351e+09\tvalid_0's l1: 67145.8\n",
      "[542]\tvalid_0's l2: 5.90365e+09\tvalid_0's l1: 67146.6\n",
      "[543]\tvalid_0's l2: 5.90374e+09\tvalid_0's l1: 67147\n",
      "[544]\tvalid_0's l2: 5.9038e+09\tvalid_0's l1: 67147\n",
      "[545]\tvalid_0's l2: 5.90382e+09\tvalid_0's l1: 67146.7\n",
      "[546]\tvalid_0's l2: 5.90377e+09\tvalid_0's l1: 67146.1\n",
      "[547]\tvalid_0's l2: 5.90374e+09\tvalid_0's l1: 67146\n",
      "[548]\tvalid_0's l2: 5.90373e+09\tvalid_0's l1: 67145.9\n",
      "[549]\tvalid_0's l2: 5.90369e+09\tvalid_0's l1: 67145.4\n",
      "[550]\tvalid_0's l2: 5.90371e+09\tvalid_0's l1: 67145.7\n",
      "[551]\tvalid_0's l2: 5.90368e+09\tvalid_0's l1: 67145.6\n",
      "[552]\tvalid_0's l2: 5.9037e+09\tvalid_0's l1: 67146\n",
      "[553]\tvalid_0's l2: 5.90373e+09\tvalid_0's l1: 67146\n",
      "[554]\tvalid_0's l2: 5.90371e+09\tvalid_0's l1: 67146.2\n",
      "[555]\tvalid_0's l2: 5.90382e+09\tvalid_0's l1: 67146.7\n",
      "[556]\tvalid_0's l2: 5.90381e+09\tvalid_0's l1: 67146.5\n",
      "[557]\tvalid_0's l2: 5.90383e+09\tvalid_0's l1: 67146.7\n",
      "[558]\tvalid_0's l2: 5.90376e+09\tvalid_0's l1: 67146.5\n",
      "[559]\tvalid_0's l2: 5.90366e+09\tvalid_0's l1: 67145.5\n",
      "[560]\tvalid_0's l2: 5.90367e+09\tvalid_0's l1: 67145.6\n",
      "[561]\tvalid_0's l2: 5.90363e+09\tvalid_0's l1: 67145\n",
      "[562]\tvalid_0's l2: 5.9037e+09\tvalid_0's l1: 67145.2\n",
      "[563]\tvalid_0's l2: 5.90381e+09\tvalid_0's l1: 67145.1\n",
      "[564]\tvalid_0's l2: 5.90387e+09\tvalid_0's l1: 67145.6\n",
      "[565]\tvalid_0's l2: 5.90391e+09\tvalid_0's l1: 67145.4\n",
      "[566]\tvalid_0's l2: 5.90394e+09\tvalid_0's l1: 67145.7\n",
      "[567]\tvalid_0's l2: 5.90386e+09\tvalid_0's l1: 67145.4\n",
      "[568]\tvalid_0's l2: 5.90377e+09\tvalid_0's l1: 67145\n",
      "[569]\tvalid_0's l2: 5.90378e+09\tvalid_0's l1: 67145.1\n",
      "[570]\tvalid_0's l2: 5.90388e+09\tvalid_0's l1: 67145.8\n",
      "[571]\tvalid_0's l2: 5.90395e+09\tvalid_0's l1: 67145.9\n",
      "[572]\tvalid_0's l2: 5.90397e+09\tvalid_0's l1: 67146.4\n",
      "[573]\tvalid_0's l2: 5.90411e+09\tvalid_0's l1: 67146.5\n",
      "[574]\tvalid_0's l2: 5.90417e+09\tvalid_0's l1: 67146.5\n",
      "[575]\tvalid_0's l2: 5.9042e+09\tvalid_0's l1: 67146.9\n",
      "[576]\tvalid_0's l2: 5.90413e+09\tvalid_0's l1: 67146.5\n",
      "[577]\tvalid_0's l2: 5.90408e+09\tvalid_0's l1: 67146.3\n",
      "[578]\tvalid_0's l2: 5.90406e+09\tvalid_0's l1: 67146.4\n",
      "[579]\tvalid_0's l2: 5.9041e+09\tvalid_0's l1: 67146.8\n",
      "[580]\tvalid_0's l2: 5.90409e+09\tvalid_0's l1: 67146.7\n",
      "[581]\tvalid_0's l2: 5.90413e+09\tvalid_0's l1: 67147.3\n",
      "[582]\tvalid_0's l2: 5.90422e+09\tvalid_0's l1: 67148\n",
      "[583]\tvalid_0's l2: 5.90426e+09\tvalid_0's l1: 67147.9\n",
      "[584]\tvalid_0's l2: 5.90425e+09\tvalid_0's l1: 67147.8\n",
      "[585]\tvalid_0's l2: 5.90428e+09\tvalid_0's l1: 67147.7\n",
      "[586]\tvalid_0's l2: 5.9043e+09\tvalid_0's l1: 67147.6\n",
      "[587]\tvalid_0's l2: 5.9043e+09\tvalid_0's l1: 67147.2\n",
      "[588]\tvalid_0's l2: 5.9042e+09\tvalid_0's l1: 67146.1\n",
      "[589]\tvalid_0's l2: 5.90421e+09\tvalid_0's l1: 67145.9\n",
      "[590]\tvalid_0's l2: 5.90412e+09\tvalid_0's l1: 67145.6\n",
      "[591]\tvalid_0's l2: 5.90405e+09\tvalid_0's l1: 67145.4\n",
      "[592]\tvalid_0's l2: 5.90398e+09\tvalid_0's l1: 67145.1\n",
      "[593]\tvalid_0's l2: 5.904e+09\tvalid_0's l1: 67145.2\n",
      "[594]\tvalid_0's l2: 5.90406e+09\tvalid_0's l1: 67145.4\n",
      "[595]\tvalid_0's l2: 5.90404e+09\tvalid_0's l1: 67145.3\n",
      "[596]\tvalid_0's l2: 5.90412e+09\tvalid_0's l1: 67146\n",
      "[597]\tvalid_0's l2: 5.90409e+09\tvalid_0's l1: 67145.9\n",
      "[598]\tvalid_0's l2: 5.90416e+09\tvalid_0's l1: 67146.3\n",
      "[599]\tvalid_0's l2: 5.90402e+09\tvalid_0's l1: 67145.6\n",
      "[600]\tvalid_0's l2: 5.90406e+09\tvalid_0's l1: 67145.6\n",
      "[601]\tvalid_0's l2: 5.90406e+09\tvalid_0's l1: 67145.5\n",
      "[602]\tvalid_0's l2: 5.90409e+09\tvalid_0's l1: 67145.7\n",
      "[603]\tvalid_0's l2: 5.90397e+09\tvalid_0's l1: 67145.1\n",
      "[604]\tvalid_0's l2: 5.90394e+09\tvalid_0's l1: 67144.9\n",
      "[605]\tvalid_0's l2: 5.90387e+09\tvalid_0's l1: 67144.8\n",
      "[606]\tvalid_0's l2: 5.90383e+09\tvalid_0's l1: 67144.3\n",
      "[607]\tvalid_0's l2: 5.90379e+09\tvalid_0's l1: 67143.9\n",
      "[608]\tvalid_0's l2: 5.90377e+09\tvalid_0's l1: 67143.7\n",
      "[609]\tvalid_0's l2: 5.90381e+09\tvalid_0's l1: 67143.9\n",
      "[610]\tvalid_0's l2: 5.90376e+09\tvalid_0's l1: 67143.7\n",
      "[611]\tvalid_0's l2: 5.90375e+09\tvalid_0's l1: 67143\n",
      "[612]\tvalid_0's l2: 5.90365e+09\tvalid_0's l1: 67142.1\n",
      "[613]\tvalid_0's l2: 5.90353e+09\tvalid_0's l1: 67141\n",
      "[614]\tvalid_0's l2: 5.90353e+09\tvalid_0's l1: 67140.6\n",
      "[615]\tvalid_0's l2: 5.9034e+09\tvalid_0's l1: 67139.9\n",
      "[616]\tvalid_0's l2: 5.90344e+09\tvalid_0's l1: 67140.5\n",
      "[617]\tvalid_0's l2: 5.90345e+09\tvalid_0's l1: 67140.4\n",
      "[618]\tvalid_0's l2: 5.90355e+09\tvalid_0's l1: 67141.1\n",
      "[619]\tvalid_0's l2: 5.90362e+09\tvalid_0's l1: 67141.5\n",
      "[620]\tvalid_0's l2: 5.90368e+09\tvalid_0's l1: 67141.8\n",
      "[621]\tvalid_0's l2: 5.90376e+09\tvalid_0's l1: 67142.5\n",
      "[622]\tvalid_0's l2: 5.90373e+09\tvalid_0's l1: 67142.4\n",
      "[623]\tvalid_0's l2: 5.90363e+09\tvalid_0's l1: 67141.9\n",
      "[624]\tvalid_0's l2: 5.90363e+09\tvalid_0's l1: 67142.1\n",
      "[625]\tvalid_0's l2: 5.90372e+09\tvalid_0's l1: 67142.8\n",
      "[626]\tvalid_0's l2: 5.90373e+09\tvalid_0's l1: 67142.9\n",
      "[627]\tvalid_0's l2: 5.9037e+09\tvalid_0's l1: 67143\n",
      "[628]\tvalid_0's l2: 5.90369e+09\tvalid_0's l1: 67142.5\n",
      "[629]\tvalid_0's l2: 5.90369e+09\tvalid_0's l1: 67142.1\n",
      "[630]\tvalid_0's l2: 5.90357e+09\tvalid_0's l1: 67141\n",
      "[631]\tvalid_0's l2: 5.90356e+09\tvalid_0's l1: 67141\n",
      "[632]\tvalid_0's l2: 5.9035e+09\tvalid_0's l1: 67141.2\n",
      "[633]\tvalid_0's l2: 5.90345e+09\tvalid_0's l1: 67141.1\n",
      "[634]\tvalid_0's l2: 5.90359e+09\tvalid_0's l1: 67141.9\n",
      "[635]\tvalid_0's l2: 5.90363e+09\tvalid_0's l1: 67142.4\n",
      "[636]\tvalid_0's l2: 5.90366e+09\tvalid_0's l1: 67142.6\n",
      "[637]\tvalid_0's l2: 5.90364e+09\tvalid_0's l1: 67142.6\n",
      "[638]\tvalid_0's l2: 5.90362e+09\tvalid_0's l1: 67142.3\n",
      "[639]\tvalid_0's l2: 5.90358e+09\tvalid_0's l1: 67142.1\n",
      "[640]\tvalid_0's l2: 5.90351e+09\tvalid_0's l1: 67141.7\n",
      "[641]\tvalid_0's l2: 5.90355e+09\tvalid_0's l1: 67141.7\n",
      "[642]\tvalid_0's l2: 5.90357e+09\tvalid_0's l1: 67141.6\n",
      "[643]\tvalid_0's l2: 5.90358e+09\tvalid_0's l1: 67141.6\n",
      "[644]\tvalid_0's l2: 5.90363e+09\tvalid_0's l1: 67141.6\n",
      "[645]\tvalid_0's l2: 5.90358e+09\tvalid_0's l1: 67141.2\n",
      "[646]\tvalid_0's l2: 5.90355e+09\tvalid_0's l1: 67141.4\n",
      "[647]\tvalid_0's l2: 5.9037e+09\tvalid_0's l1: 67141.9\n",
      "[648]\tvalid_0's l2: 5.9036e+09\tvalid_0's l1: 67141.5\n",
      "[649]\tvalid_0's l2: 5.90357e+09\tvalid_0's l1: 67141.1\n",
      "[650]\tvalid_0's l2: 5.90357e+09\tvalid_0's l1: 67141.3\n",
      "[651]\tvalid_0's l2: 5.90377e+09\tvalid_0's l1: 67142.4\n",
      "[652]\tvalid_0's l2: 5.90372e+09\tvalid_0's l1: 67142.2\n",
      "[653]\tvalid_0's l2: 5.9038e+09\tvalid_0's l1: 67142.7\n",
      "[654]\tvalid_0's l2: 5.90378e+09\tvalid_0's l1: 67142.9\n",
      "[655]\tvalid_0's l2: 5.90375e+09\tvalid_0's l1: 67143\n",
      "[656]\tvalid_0's l2: 5.90373e+09\tvalid_0's l1: 67143.3\n",
      "[657]\tvalid_0's l2: 5.90366e+09\tvalid_0's l1: 67142.9\n",
      "[658]\tvalid_0's l2: 5.90369e+09\tvalid_0's l1: 67143.3\n",
      "[659]\tvalid_0's l2: 5.90374e+09\tvalid_0's l1: 67143.6\n",
      "[660]\tvalid_0's l2: 5.90374e+09\tvalid_0's l1: 67143.8\n",
      "[661]\tvalid_0's l2: 5.90376e+09\tvalid_0's l1: 67143.8\n",
      "[662]\tvalid_0's l2: 5.90377e+09\tvalid_0's l1: 67143.7\n",
      "[663]\tvalid_0's l2: 5.90367e+09\tvalid_0's l1: 67143.2\n",
      "[664]\tvalid_0's l2: 5.9037e+09\tvalid_0's l1: 67143.2\n",
      "[665]\tvalid_0's l2: 5.90371e+09\tvalid_0's l1: 67143\n",
      "[666]\tvalid_0's l2: 5.9036e+09\tvalid_0's l1: 67141.7\n",
      "[667]\tvalid_0's l2: 5.90366e+09\tvalid_0's l1: 67141.9\n",
      "[668]\tvalid_0's l2: 5.90371e+09\tvalid_0's l1: 67142.2\n",
      "[669]\tvalid_0's l2: 5.90375e+09\tvalid_0's l1: 67141.8\n",
      "[670]\tvalid_0's l2: 5.90373e+09\tvalid_0's l1: 67141.3\n",
      "[671]\tvalid_0's l2: 5.90363e+09\tvalid_0's l1: 67140.3\n",
      "[672]\tvalid_0's l2: 5.90351e+09\tvalid_0's l1: 67140\n",
      "[673]\tvalid_0's l2: 5.90345e+09\tvalid_0's l1: 67139.1\n",
      "[674]\tvalid_0's l2: 5.90337e+09\tvalid_0's l1: 67138.2\n",
      "[675]\tvalid_0's l2: 5.9033e+09\tvalid_0's l1: 67137.6\n",
      "[676]\tvalid_0's l2: 5.90328e+09\tvalid_0's l1: 67136.8\n",
      "[677]\tvalid_0's l2: 5.90341e+09\tvalid_0's l1: 67136.9\n",
      "[678]\tvalid_0's l2: 5.9034e+09\tvalid_0's l1: 67136.1\n",
      "[679]\tvalid_0's l2: 5.90336e+09\tvalid_0's l1: 67135.7\n",
      "[680]\tvalid_0's l2: 5.90341e+09\tvalid_0's l1: 67135.6\n",
      "[681]\tvalid_0's l2: 5.90351e+09\tvalid_0's l1: 67136.1\n",
      "[682]\tvalid_0's l2: 5.90345e+09\tvalid_0's l1: 67135.6\n",
      "[683]\tvalid_0's l2: 5.9036e+09\tvalid_0's l1: 67136.4\n",
      "[684]\tvalid_0's l2: 5.90362e+09\tvalid_0's l1: 67136.4\n",
      "[685]\tvalid_0's l2: 5.90363e+09\tvalid_0's l1: 67136.5\n",
      "[686]\tvalid_0's l2: 5.90378e+09\tvalid_0's l1: 67138.1\n",
      "[687]\tvalid_0's l2: 5.90389e+09\tvalid_0's l1: 67139.3\n",
      "[688]\tvalid_0's l2: 5.90398e+09\tvalid_0's l1: 67140.1\n",
      "[689]\tvalid_0's l2: 5.90406e+09\tvalid_0's l1: 67140.6\n",
      "[690]\tvalid_0's l2: 5.90406e+09\tvalid_0's l1: 67140.6\n",
      "[691]\tvalid_0's l2: 5.90419e+09\tvalid_0's l1: 67140.9\n",
      "[692]\tvalid_0's l2: 5.90407e+09\tvalid_0's l1: 67139.6\n",
      "[693]\tvalid_0's l2: 5.9042e+09\tvalid_0's l1: 67140\n",
      "[694]\tvalid_0's l2: 5.90415e+09\tvalid_0's l1: 67139\n",
      "[695]\tvalid_0's l2: 5.90427e+09\tvalid_0's l1: 67139.1\n",
      "[696]\tvalid_0's l2: 5.9043e+09\tvalid_0's l1: 67139.6\n",
      "[697]\tvalid_0's l2: 5.90428e+09\tvalid_0's l1: 67139.3\n",
      "[698]\tvalid_0's l2: 5.90426e+09\tvalid_0's l1: 67139.3\n",
      "[699]\tvalid_0's l2: 5.90421e+09\tvalid_0's l1: 67139.7\n",
      "[700]\tvalid_0's l2: 5.9042e+09\tvalid_0's l1: 67139.7\n",
      "[701]\tvalid_0's l2: 5.90426e+09\tvalid_0's l1: 67140.1\n",
      "[702]\tvalid_0's l2: 5.90414e+09\tvalid_0's l1: 67139.1\n",
      "[703]\tvalid_0's l2: 5.90409e+09\tvalid_0's l1: 67139\n",
      "[704]\tvalid_0's l2: 5.90407e+09\tvalid_0's l1: 67138.5\n",
      "[705]\tvalid_0's l2: 5.90395e+09\tvalid_0's l1: 67137.9\n",
      "[706]\tvalid_0's l2: 5.90393e+09\tvalid_0's l1: 67137.1\n",
      "[707]\tvalid_0's l2: 5.90399e+09\tvalid_0's l1: 67137\n",
      "[708]\tvalid_0's l2: 5.90405e+09\tvalid_0's l1: 67137\n",
      "[709]\tvalid_0's l2: 5.90417e+09\tvalid_0's l1: 67137.2\n",
      "[710]\tvalid_0's l2: 5.90428e+09\tvalid_0's l1: 67137.8\n",
      "[711]\tvalid_0's l2: 5.90437e+09\tvalid_0's l1: 67138.1\n",
      "[712]\tvalid_0's l2: 5.90434e+09\tvalid_0's l1: 67138.2\n",
      "[713]\tvalid_0's l2: 5.90435e+09\tvalid_0's l1: 67138.4\n",
      "[714]\tvalid_0's l2: 5.90428e+09\tvalid_0's l1: 67138.1\n",
      "[715]\tvalid_0's l2: 5.90429e+09\tvalid_0's l1: 67138.4\n",
      "[716]\tvalid_0's l2: 5.90426e+09\tvalid_0's l1: 67138.6\n",
      "[717]\tvalid_0's l2: 5.90424e+09\tvalid_0's l1: 67138.7\n",
      "[718]\tvalid_0's l2: 5.90425e+09\tvalid_0's l1: 67139\n",
      "[719]\tvalid_0's l2: 5.90411e+09\tvalid_0's l1: 67138.3\n",
      "[720]\tvalid_0's l2: 5.90417e+09\tvalid_0's l1: 67138.3\n",
      "[721]\tvalid_0's l2: 5.9043e+09\tvalid_0's l1: 67139.4\n",
      "[722]\tvalid_0's l2: 5.90434e+09\tvalid_0's l1: 67139.5\n",
      "[723]\tvalid_0's l2: 5.90437e+09\tvalid_0's l1: 67139.8\n",
      "[724]\tvalid_0's l2: 5.90437e+09\tvalid_0's l1: 67139.9\n",
      "[725]\tvalid_0's l2: 5.90445e+09\tvalid_0's l1: 67140.1\n",
      "[726]\tvalid_0's l2: 5.90466e+09\tvalid_0's l1: 67141.2\n",
      "[727]\tvalid_0's l2: 5.90468e+09\tvalid_0's l1: 67140.7\n",
      "[728]\tvalid_0's l2: 5.90478e+09\tvalid_0's l1: 67141.4\n",
      "[729]\tvalid_0's l2: 5.90482e+09\tvalid_0's l1: 67141.5\n",
      "[730]\tvalid_0's l2: 5.90496e+09\tvalid_0's l1: 67142.1\n",
      "[731]\tvalid_0's l2: 5.90493e+09\tvalid_0's l1: 67142\n",
      "[732]\tvalid_0's l2: 5.90497e+09\tvalid_0's l1: 67142.1\n",
      "[733]\tvalid_0's l2: 5.90501e+09\tvalid_0's l1: 67142.2\n",
      "[734]\tvalid_0's l2: 5.90497e+09\tvalid_0's l1: 67141.9\n",
      "[735]\tvalid_0's l2: 5.90497e+09\tvalid_0's l1: 67141.8\n",
      "[736]\tvalid_0's l2: 5.90515e+09\tvalid_0's l1: 67143\n",
      "[737]\tvalid_0's l2: 5.90514e+09\tvalid_0's l1: 67142.5\n",
      "[738]\tvalid_0's l2: 5.90512e+09\tvalid_0's l1: 67142.6\n",
      "[739]\tvalid_0's l2: 5.90509e+09\tvalid_0's l1: 67142.4\n",
      "[740]\tvalid_0's l2: 5.90514e+09\tvalid_0's l1: 67142.6\n",
      "[741]\tvalid_0's l2: 5.905e+09\tvalid_0's l1: 67141.9\n",
      "[742]\tvalid_0's l2: 5.90491e+09\tvalid_0's l1: 67141\n",
      "[743]\tvalid_0's l2: 5.90473e+09\tvalid_0's l1: 67140.2\n",
      "[744]\tvalid_0's l2: 5.90472e+09\tvalid_0's l1: 67140.1\n",
      "[745]\tvalid_0's l2: 5.90463e+09\tvalid_0's l1: 67139.9\n",
      "[746]\tvalid_0's l2: 5.90461e+09\tvalid_0's l1: 67139.8\n",
      "[747]\tvalid_0's l2: 5.9046e+09\tvalid_0's l1: 67139.6\n",
      "[748]\tvalid_0's l2: 5.9046e+09\tvalid_0's l1: 67139.3\n",
      "[749]\tvalid_0's l2: 5.90469e+09\tvalid_0's l1: 67139.1\n",
      "[750]\tvalid_0's l2: 5.90466e+09\tvalid_0's l1: 67138.6\n",
      "[751]\tvalid_0's l2: 5.90458e+09\tvalid_0's l1: 67138.5\n",
      "[752]\tvalid_0's l2: 5.90446e+09\tvalid_0's l1: 67138.2\n",
      "[753]\tvalid_0's l2: 5.90448e+09\tvalid_0's l1: 67138.4\n",
      "[754]\tvalid_0's l2: 5.90445e+09\tvalid_0's l1: 67138.3\n",
      "[755]\tvalid_0's l2: 5.90431e+09\tvalid_0's l1: 67137.5\n",
      "[756]\tvalid_0's l2: 5.90428e+09\tvalid_0's l1: 67137.4\n",
      "[757]\tvalid_0's l2: 5.90436e+09\tvalid_0's l1: 67137.6\n",
      "[758]\tvalid_0's l2: 5.90432e+09\tvalid_0's l1: 67137.5\n",
      "[759]\tvalid_0's l2: 5.9044e+09\tvalid_0's l1: 67137.9\n",
      "[760]\tvalid_0's l2: 5.90438e+09\tvalid_0's l1: 67137.8\n",
      "[761]\tvalid_0's l2: 5.90439e+09\tvalid_0's l1: 67137.4\n",
      "[762]\tvalid_0's l2: 5.90437e+09\tvalid_0's l1: 67136.9\n",
      "[763]\tvalid_0's l2: 5.90426e+09\tvalid_0's l1: 67136.1\n",
      "[764]\tvalid_0's l2: 5.90432e+09\tvalid_0's l1: 67136\n",
      "[765]\tvalid_0's l2: 5.90427e+09\tvalid_0's l1: 67136.3\n",
      "[766]\tvalid_0's l2: 5.90426e+09\tvalid_0's l1: 67136.1\n",
      "[767]\tvalid_0's l2: 5.90424e+09\tvalid_0's l1: 67135.4\n",
      "[768]\tvalid_0's l2: 5.90425e+09\tvalid_0's l1: 67135.8\n",
      "[769]\tvalid_0's l2: 5.90419e+09\tvalid_0's l1: 67134.9\n",
      "[770]\tvalid_0's l2: 5.90415e+09\tvalid_0's l1: 67134.4\n",
      "[771]\tvalid_0's l2: 5.90419e+09\tvalid_0's l1: 67134.7\n",
      "[772]\tvalid_0's l2: 5.90423e+09\tvalid_0's l1: 67134.6\n",
      "[773]\tvalid_0's l2: 5.90428e+09\tvalid_0's l1: 67134.8\n",
      "[774]\tvalid_0's l2: 5.90442e+09\tvalid_0's l1: 67135.4\n",
      "[775]\tvalid_0's l2: 5.90438e+09\tvalid_0's l1: 67134.9\n",
      "[776]\tvalid_0's l2: 5.90432e+09\tvalid_0's l1: 67133.9\n",
      "[777]\tvalid_0's l2: 5.90436e+09\tvalid_0's l1: 67133.6\n",
      "[778]\tvalid_0's l2: 5.9043e+09\tvalid_0's l1: 67132.9\n",
      "[779]\tvalid_0's l2: 5.90425e+09\tvalid_0's l1: 67132.2\n",
      "[780]\tvalid_0's l2: 5.90419e+09\tvalid_0's l1: 67131.6\n",
      "[781]\tvalid_0's l2: 5.90424e+09\tvalid_0's l1: 67132.2\n",
      "[782]\tvalid_0's l2: 5.90425e+09\tvalid_0's l1: 67132.8\n",
      "[783]\tvalid_0's l2: 5.90431e+09\tvalid_0's l1: 67133.3\n",
      "[784]\tvalid_0's l2: 5.90437e+09\tvalid_0's l1: 67133.9\n",
      "[785]\tvalid_0's l2: 5.90441e+09\tvalid_0's l1: 67134\n",
      "[786]\tvalid_0's l2: 5.90439e+09\tvalid_0's l1: 67134.2\n",
      "[787]\tvalid_0's l2: 5.90429e+09\tvalid_0's l1: 67133.4\n",
      "[788]\tvalid_0's l2: 5.90426e+09\tvalid_0's l1: 67133.4\n",
      "[789]\tvalid_0's l2: 5.90438e+09\tvalid_0's l1: 67134.1\n",
      "[790]\tvalid_0's l2: 5.90443e+09\tvalid_0's l1: 67134.7\n",
      "[791]\tvalid_0's l2: 5.90432e+09\tvalid_0's l1: 67134.3\n",
      "[792]\tvalid_0's l2: 5.90437e+09\tvalid_0's l1: 67134.7\n",
      "[793]\tvalid_0's l2: 5.90433e+09\tvalid_0's l1: 67134.9\n",
      "[794]\tvalid_0's l2: 5.90433e+09\tvalid_0's l1: 67135.1\n",
      "[795]\tvalid_0's l2: 5.90441e+09\tvalid_0's l1: 67135.8\n",
      "[796]\tvalid_0's l2: 5.90439e+09\tvalid_0's l1: 67135.4\n",
      "[797]\tvalid_0's l2: 5.90428e+09\tvalid_0's l1: 67134.5\n",
      "[798]\tvalid_0's l2: 5.90429e+09\tvalid_0's l1: 67134.5\n",
      "[799]\tvalid_0's l2: 5.90438e+09\tvalid_0's l1: 67134.4\n",
      "[800]\tvalid_0's l2: 5.90444e+09\tvalid_0's l1: 67134.4\n",
      "[801]\tvalid_0's l2: 5.90457e+09\tvalid_0's l1: 67135.2\n",
      "[802]\tvalid_0's l2: 5.90469e+09\tvalid_0's l1: 67135.8\n",
      "[803]\tvalid_0's l2: 5.90472e+09\tvalid_0's l1: 67135.8\n",
      "[804]\tvalid_0's l2: 5.90467e+09\tvalid_0's l1: 67135.9\n",
      "[805]\tvalid_0's l2: 5.90488e+09\tvalid_0's l1: 67137.3\n",
      "[806]\tvalid_0's l2: 5.90499e+09\tvalid_0's l1: 67137.7\n",
      "[807]\tvalid_0's l2: 5.9051e+09\tvalid_0's l1: 67138.2\n",
      "[808]\tvalid_0's l2: 5.90501e+09\tvalid_0's l1: 67137.6\n",
      "[809]\tvalid_0's l2: 5.90512e+09\tvalid_0's l1: 67138\n",
      "[810]\tvalid_0's l2: 5.90508e+09\tvalid_0's l1: 67137.3\n",
      "[811]\tvalid_0's l2: 5.90511e+09\tvalid_0's l1: 67137.5\n",
      "[812]\tvalid_0's l2: 5.90514e+09\tvalid_0's l1: 67137.6\n",
      "[813]\tvalid_0's l2: 5.90518e+09\tvalid_0's l1: 67137.7\n",
      "[814]\tvalid_0's l2: 5.9051e+09\tvalid_0's l1: 67137.4\n",
      "[815]\tvalid_0's l2: 5.90519e+09\tvalid_0's l1: 67138.1\n",
      "[816]\tvalid_0's l2: 5.90539e+09\tvalid_0's l1: 67139.6\n",
      "[817]\tvalid_0's l2: 5.90546e+09\tvalid_0's l1: 67140.1\n",
      "[818]\tvalid_0's l2: 5.90548e+09\tvalid_0's l1: 67140.2\n",
      "[819]\tvalid_0's l2: 5.90547e+09\tvalid_0's l1: 67140.3\n",
      "[820]\tvalid_0's l2: 5.90543e+09\tvalid_0's l1: 67140.1\n",
      "[821]\tvalid_0's l2: 5.90534e+09\tvalid_0's l1: 67139.6\n",
      "[822]\tvalid_0's l2: 5.90536e+09\tvalid_0's l1: 67139.9\n",
      "[823]\tvalid_0's l2: 5.90539e+09\tvalid_0's l1: 67140.2\n",
      "[824]\tvalid_0's l2: 5.90521e+09\tvalid_0's l1: 67138.9\n",
      "[825]\tvalid_0's l2: 5.90528e+09\tvalid_0's l1: 67139.4\n",
      "[826]\tvalid_0's l2: 5.90537e+09\tvalid_0's l1: 67139.7\n",
      "[827]\tvalid_0's l2: 5.90529e+09\tvalid_0's l1: 67139\n",
      "[828]\tvalid_0's l2: 5.90506e+09\tvalid_0's l1: 67137.1\n",
      "[829]\tvalid_0's l2: 5.90505e+09\tvalid_0's l1: 67136.9\n",
      "[830]\tvalid_0's l2: 5.90512e+09\tvalid_0's l1: 67136.9\n",
      "[831]\tvalid_0's l2: 5.90501e+09\tvalid_0's l1: 67136.2\n",
      "[832]\tvalid_0's l2: 5.90488e+09\tvalid_0's l1: 67135.2\n",
      "[833]\tvalid_0's l2: 5.90489e+09\tvalid_0's l1: 67135.5\n",
      "[834]\tvalid_0's l2: 5.90489e+09\tvalid_0's l1: 67135.1\n",
      "[835]\tvalid_0's l2: 5.90502e+09\tvalid_0's l1: 67135.2\n",
      "[836]\tvalid_0's l2: 5.90503e+09\tvalid_0's l1: 67135.1\n",
      "[837]\tvalid_0's l2: 5.90511e+09\tvalid_0's l1: 67135.3\n",
      "[838]\tvalid_0's l2: 5.90515e+09\tvalid_0's l1: 67135.3\n",
      "[839]\tvalid_0's l2: 5.90515e+09\tvalid_0's l1: 67135.2\n",
      "[840]\tvalid_0's l2: 5.90527e+09\tvalid_0's l1: 67135.8\n",
      "[841]\tvalid_0's l2: 5.9053e+09\tvalid_0's l1: 67135.7\n",
      "[842]\tvalid_0's l2: 5.90534e+09\tvalid_0's l1: 67135.5\n",
      "[843]\tvalid_0's l2: 5.90537e+09\tvalid_0's l1: 67135.4\n",
      "[844]\tvalid_0's l2: 5.90538e+09\tvalid_0's l1: 67135.4\n",
      "[845]\tvalid_0's l2: 5.90529e+09\tvalid_0's l1: 67134.8\n",
      "[846]\tvalid_0's l2: 5.90518e+09\tvalid_0's l1: 67133.9\n",
      "[847]\tvalid_0's l2: 5.90534e+09\tvalid_0's l1: 67135\n",
      "[848]\tvalid_0's l2: 5.9053e+09\tvalid_0's l1: 67134.5\n",
      "[849]\tvalid_0's l2: 5.90544e+09\tvalid_0's l1: 67135.4\n",
      "[850]\tvalid_0's l2: 5.90541e+09\tvalid_0's l1: 67134.9\n",
      "[851]\tvalid_0's l2: 5.90543e+09\tvalid_0's l1: 67135\n",
      "[852]\tvalid_0's l2: 5.9055e+09\tvalid_0's l1: 67135.3\n",
      "[853]\tvalid_0's l2: 5.9055e+09\tvalid_0's l1: 67135\n",
      "[854]\tvalid_0's l2: 5.90561e+09\tvalid_0's l1: 67136\n",
      "[855]\tvalid_0's l2: 5.90569e+09\tvalid_0's l1: 67136.5\n",
      "[856]\tvalid_0's l2: 5.90581e+09\tvalid_0's l1: 67137.5\n",
      "[857]\tvalid_0's l2: 5.9059e+09\tvalid_0's l1: 67137.9\n",
      "[858]\tvalid_0's l2: 5.90584e+09\tvalid_0's l1: 67137.3\n",
      "[859]\tvalid_0's l2: 5.90588e+09\tvalid_0's l1: 67137.2\n",
      "[860]\tvalid_0's l2: 5.90599e+09\tvalid_0's l1: 67137.5\n",
      "[861]\tvalid_0's l2: 5.90601e+09\tvalid_0's l1: 67137.3\n",
      "[862]\tvalid_0's l2: 5.90603e+09\tvalid_0's l1: 67137.2\n",
      "[863]\tvalid_0's l2: 5.90606e+09\tvalid_0's l1: 67137.1\n",
      "[864]\tvalid_0's l2: 5.90609e+09\tvalid_0's l1: 67136.8\n",
      "[865]\tvalid_0's l2: 5.90615e+09\tvalid_0's l1: 67137\n",
      "[866]\tvalid_0's l2: 5.90614e+09\tvalid_0's l1: 67136.9\n",
      "[867]\tvalid_0's l2: 5.90607e+09\tvalid_0's l1: 67136.4\n",
      "[868]\tvalid_0's l2: 5.90608e+09\tvalid_0's l1: 67136.3\n",
      "[869]\tvalid_0's l2: 5.90606e+09\tvalid_0's l1: 67135.7\n",
      "[870]\tvalid_0's l2: 5.90599e+09\tvalid_0's l1: 67135.1\n",
      "[871]\tvalid_0's l2: 5.90591e+09\tvalid_0's l1: 67134.8\n",
      "[872]\tvalid_0's l2: 5.90584e+09\tvalid_0's l1: 67134.5\n",
      "[873]\tvalid_0's l2: 5.90563e+09\tvalid_0's l1: 67133.2\n",
      "[874]\tvalid_0's l2: 5.9056e+09\tvalid_0's l1: 67133.1\n",
      "[875]\tvalid_0's l2: 5.90547e+09\tvalid_0's l1: 67132.3\n",
      "[876]\tvalid_0's l2: 5.90548e+09\tvalid_0's l1: 67132.1\n",
      "[877]\tvalid_0's l2: 5.90556e+09\tvalid_0's l1: 67132.3\n",
      "[878]\tvalid_0's l2: 5.9054e+09\tvalid_0's l1: 67131.2\n",
      "[879]\tvalid_0's l2: 5.90535e+09\tvalid_0's l1: 67130.6\n",
      "[880]\tvalid_0's l2: 5.90548e+09\tvalid_0's l1: 67131.6\n",
      "[881]\tvalid_0's l2: 5.90557e+09\tvalid_0's l1: 67132.4\n",
      "[882]\tvalid_0's l2: 5.90551e+09\tvalid_0's l1: 67132\n",
      "[883]\tvalid_0's l2: 5.90555e+09\tvalid_0's l1: 67132.5\n",
      "[884]\tvalid_0's l2: 5.90565e+09\tvalid_0's l1: 67133\n",
      "[885]\tvalid_0's l2: 5.90544e+09\tvalid_0's l1: 67132\n",
      "[886]\tvalid_0's l2: 5.90547e+09\tvalid_0's l1: 67132\n",
      "[887]\tvalid_0's l2: 5.90539e+09\tvalid_0's l1: 67131.5\n",
      "[888]\tvalid_0's l2: 5.90538e+09\tvalid_0's l1: 67131.3\n",
      "[889]\tvalid_0's l2: 5.90523e+09\tvalid_0's l1: 67129.9\n",
      "[890]\tvalid_0's l2: 5.90516e+09\tvalid_0's l1: 67129.3\n",
      "[891]\tvalid_0's l2: 5.90517e+09\tvalid_0's l1: 67129.3\n",
      "[892]\tvalid_0's l2: 5.90521e+09\tvalid_0's l1: 67129.6\n",
      "[893]\tvalid_0's l2: 5.90525e+09\tvalid_0's l1: 67129.9\n",
      "[894]\tvalid_0's l2: 5.90528e+09\tvalid_0's l1: 67129.9\n",
      "[895]\tvalid_0's l2: 5.90523e+09\tvalid_0's l1: 67129.6\n",
      "[896]\tvalid_0's l2: 5.90526e+09\tvalid_0's l1: 67129.5\n",
      "[897]\tvalid_0's l2: 5.90537e+09\tvalid_0's l1: 67130.8\n",
      "[898]\tvalid_0's l2: 5.90525e+09\tvalid_0's l1: 67130.3\n",
      "[899]\tvalid_0's l2: 5.90539e+09\tvalid_0's l1: 67131.4\n",
      "[900]\tvalid_0's l2: 5.90549e+09\tvalid_0's l1: 67131.9\n",
      "[901]\tvalid_0's l2: 5.90546e+09\tvalid_0's l1: 67131.8\n",
      "[902]\tvalid_0's l2: 5.90555e+09\tvalid_0's l1: 67131.8\n",
      "[903]\tvalid_0's l2: 5.90562e+09\tvalid_0's l1: 67132.3\n",
      "[904]\tvalid_0's l2: 5.90558e+09\tvalid_0's l1: 67131.9\n",
      "[905]\tvalid_0's l2: 5.90557e+09\tvalid_0's l1: 67131.1\n",
      "[906]\tvalid_0's l2: 5.90555e+09\tvalid_0's l1: 67131.1\n",
      "[907]\tvalid_0's l2: 5.90554e+09\tvalid_0's l1: 67131.1\n",
      "[908]\tvalid_0's l2: 5.90554e+09\tvalid_0's l1: 67131.1\n",
      "[909]\tvalid_0's l2: 5.90553e+09\tvalid_0's l1: 67131.1\n",
      "[910]\tvalid_0's l2: 5.90557e+09\tvalid_0's l1: 67131.7\n",
      "[911]\tvalid_0's l2: 5.90555e+09\tvalid_0's l1: 67131.3\n",
      "[912]\tvalid_0's l2: 5.90568e+09\tvalid_0's l1: 67131.9\n",
      "[913]\tvalid_0's l2: 5.90582e+09\tvalid_0's l1: 67132.7\n",
      "[914]\tvalid_0's l2: 5.90596e+09\tvalid_0's l1: 67133.4\n",
      "[915]\tvalid_0's l2: 5.90596e+09\tvalid_0's l1: 67133.5\n",
      "[916]\tvalid_0's l2: 5.90596e+09\tvalid_0's l1: 67133.4\n",
      "[917]\tvalid_0's l2: 5.90599e+09\tvalid_0's l1: 67133.3\n",
      "[918]\tvalid_0's l2: 5.90591e+09\tvalid_0's l1: 67132.7\n",
      "[919]\tvalid_0's l2: 5.906e+09\tvalid_0's l1: 67133.4\n",
      "[920]\tvalid_0's l2: 5.90606e+09\tvalid_0's l1: 67133\n",
      "[921]\tvalid_0's l2: 5.90606e+09\tvalid_0's l1: 67133\n",
      "[922]\tvalid_0's l2: 5.90604e+09\tvalid_0's l1: 67132.9\n",
      "[923]\tvalid_0's l2: 5.90598e+09\tvalid_0's l1: 67132.7\n",
      "[924]\tvalid_0's l2: 5.90595e+09\tvalid_0's l1: 67132.4\n",
      "[925]\tvalid_0's l2: 5.90594e+09\tvalid_0's l1: 67132\n",
      "[926]\tvalid_0's l2: 5.90596e+09\tvalid_0's l1: 67132.1\n",
      "[927]\tvalid_0's l2: 5.90603e+09\tvalid_0's l1: 67132.2\n",
      "[928]\tvalid_0's l2: 5.90605e+09\tvalid_0's l1: 67132.3\n",
      "[929]\tvalid_0's l2: 5.90602e+09\tvalid_0's l1: 67132.2\n",
      "[930]\tvalid_0's l2: 5.90599e+09\tvalid_0's l1: 67131.6\n",
      "[931]\tvalid_0's l2: 5.90614e+09\tvalid_0's l1: 67132.1\n",
      "[932]\tvalid_0's l2: 5.90611e+09\tvalid_0's l1: 67131.7\n",
      "[933]\tvalid_0's l2: 5.90598e+09\tvalid_0's l1: 67130.9\n",
      "[934]\tvalid_0's l2: 5.90597e+09\tvalid_0's l1: 67130.6\n",
      "[935]\tvalid_0's l2: 5.90616e+09\tvalid_0's l1: 67131.4\n",
      "[936]\tvalid_0's l2: 5.90619e+09\tvalid_0's l1: 67131.5\n",
      "[937]\tvalid_0's l2: 5.90625e+09\tvalid_0's l1: 67131.7\n",
      "[938]\tvalid_0's l2: 5.90623e+09\tvalid_0's l1: 67131.6\n",
      "[939]\tvalid_0's l2: 5.90625e+09\tvalid_0's l1: 67131.8\n",
      "[940]\tvalid_0's l2: 5.90616e+09\tvalid_0's l1: 67130.7\n",
      "[941]\tvalid_0's l2: 5.9062e+09\tvalid_0's l1: 67130.3\n",
      "[942]\tvalid_0's l2: 5.9063e+09\tvalid_0's l1: 67130.4\n",
      "[943]\tvalid_0's l2: 5.90622e+09\tvalid_0's l1: 67130.1\n",
      "[944]\tvalid_0's l2: 5.90622e+09\tvalid_0's l1: 67129.8\n",
      "[945]\tvalid_0's l2: 5.90621e+09\tvalid_0's l1: 67129.4\n",
      "[946]\tvalid_0's l2: 5.90619e+09\tvalid_0's l1: 67129.4\n",
      "[947]\tvalid_0's l2: 5.90612e+09\tvalid_0's l1: 67129.1\n",
      "[948]\tvalid_0's l2: 5.90614e+09\tvalid_0's l1: 67129.3\n",
      "[949]\tvalid_0's l2: 5.90615e+09\tvalid_0's l1: 67129.3\n",
      "[950]\tvalid_0's l2: 5.90627e+09\tvalid_0's l1: 67130\n",
      "[951]\tvalid_0's l2: 5.90634e+09\tvalid_0's l1: 67130.4\n",
      "[952]\tvalid_0's l2: 5.90631e+09\tvalid_0's l1: 67129.9\n",
      "[953]\tvalid_0's l2: 5.90604e+09\tvalid_0's l1: 67128.5\n",
      "[954]\tvalid_0's l2: 5.90609e+09\tvalid_0's l1: 67128.8\n",
      "[955]\tvalid_0's l2: 5.90626e+09\tvalid_0's l1: 67129.6\n",
      "[956]\tvalid_0's l2: 5.90627e+09\tvalid_0's l1: 67129.7\n",
      "[957]\tvalid_0's l2: 5.90627e+09\tvalid_0's l1: 67129.6\n",
      "[958]\tvalid_0's l2: 5.90631e+09\tvalid_0's l1: 67130\n",
      "[959]\tvalid_0's l2: 5.9064e+09\tvalid_0's l1: 67130.2\n",
      "[960]\tvalid_0's l2: 5.90644e+09\tvalid_0's l1: 67130.5\n",
      "[961]\tvalid_0's l2: 5.90645e+09\tvalid_0's l1: 67130.7\n",
      "[962]\tvalid_0's l2: 5.90642e+09\tvalid_0's l1: 67130.9\n",
      "[963]\tvalid_0's l2: 5.90653e+09\tvalid_0's l1: 67131.5\n",
      "[964]\tvalid_0's l2: 5.90655e+09\tvalid_0's l1: 67131.8\n",
      "[965]\tvalid_0's l2: 5.9067e+09\tvalid_0's l1: 67132.7\n",
      "[966]\tvalid_0's l2: 5.90656e+09\tvalid_0's l1: 67131.7\n",
      "[967]\tvalid_0's l2: 5.90634e+09\tvalid_0's l1: 67130.7\n",
      "[968]\tvalid_0's l2: 5.90624e+09\tvalid_0's l1: 67130\n",
      "[969]\tvalid_0's l2: 5.9062e+09\tvalid_0's l1: 67129.7\n",
      "[970]\tvalid_0's l2: 5.90604e+09\tvalid_0's l1: 67128.6\n",
      "[971]\tvalid_0's l2: 5.90599e+09\tvalid_0's l1: 67127.9\n",
      "[972]\tvalid_0's l2: 5.90596e+09\tvalid_0's l1: 67127.4\n",
      "[973]\tvalid_0's l2: 5.90596e+09\tvalid_0's l1: 67127.3\n",
      "[974]\tvalid_0's l2: 5.90597e+09\tvalid_0's l1: 67127.3\n",
      "[975]\tvalid_0's l2: 5.90581e+09\tvalid_0's l1: 67126.2\n",
      "[976]\tvalid_0's l2: 5.90582e+09\tvalid_0's l1: 67126.1\n",
      "[977]\tvalid_0's l2: 5.90584e+09\tvalid_0's l1: 67126.5\n",
      "[978]\tvalid_0's l2: 5.90579e+09\tvalid_0's l1: 67126.3\n",
      "[979]\tvalid_0's l2: 5.90582e+09\tvalid_0's l1: 67126.5\n",
      "[980]\tvalid_0's l2: 5.90586e+09\tvalid_0's l1: 67126.8\n",
      "[981]\tvalid_0's l2: 5.90573e+09\tvalid_0's l1: 67125.9\n",
      "[982]\tvalid_0's l2: 5.90583e+09\tvalid_0's l1: 67126.6\n",
      "[983]\tvalid_0's l2: 5.90584e+09\tvalid_0's l1: 67126.5\n",
      "[984]\tvalid_0's l2: 5.90574e+09\tvalid_0's l1: 67126.1\n",
      "[985]\tvalid_0's l2: 5.90578e+09\tvalid_0's l1: 67126.3\n",
      "[986]\tvalid_0's l2: 5.90589e+09\tvalid_0's l1: 67127.1\n",
      "[987]\tvalid_0's l2: 5.90577e+09\tvalid_0's l1: 67125.9\n",
      "[988]\tvalid_0's l2: 5.9057e+09\tvalid_0's l1: 67125\n",
      "[989]\tvalid_0's l2: 5.90585e+09\tvalid_0's l1: 67126.2\n",
      "[990]\tvalid_0's l2: 5.90584e+09\tvalid_0's l1: 67125.6\n",
      "[991]\tvalid_0's l2: 5.90575e+09\tvalid_0's l1: 67125.3\n",
      "[992]\tvalid_0's l2: 5.90567e+09\tvalid_0's l1: 67125\n",
      "[993]\tvalid_0's l2: 5.90564e+09\tvalid_0's l1: 67125.1\n",
      "[994]\tvalid_0's l2: 5.90563e+09\tvalid_0's l1: 67125\n",
      "[995]\tvalid_0's l2: 5.90567e+09\tvalid_0's l1: 67124.9\n",
      "[996]\tvalid_0's l2: 5.90566e+09\tvalid_0's l1: 67124.9\n",
      "[997]\tvalid_0's l2: 5.90565e+09\tvalid_0's l1: 67124.7\n",
      "[998]\tvalid_0's l2: 5.90568e+09\tvalid_0's l1: 67125.1\n",
      "[999]\tvalid_0's l2: 5.90558e+09\tvalid_0's l1: 67124.2\n",
      "[1000]\tvalid_0's l2: 5.90556e+09\tvalid_0's l1: 67124.2\n",
      "The rmse of prediction is: 76847.61354285045\n"
     ]
    }
   ],
   "source": [
    "# create dataset for lightgbm\n",
    "lgb_train = lgb.Dataset(X_train_inst, y_train_inst)\n",
    "lgb_eval = lgb.Dataset(X_test_inst, y_test_inst, reference=lgb_train)\n",
    "\n",
    "params = {\n",
    "    'n_estimators' : 1000,\n",
    "    'boosting_type': 'gbdt',\n",
    "    #'boosting_type': 'dart',\n",
    "    'objective': 'regression',\n",
    "    'metric': {'l2', 'l1'},\n",
    "    'num_leaves': 6,\n",
    "    'learning_rate': 0.05,\n",
    "    'feature_fraction': 0.9,\n",
    "    'bagging_fraction': 0.8,\n",
    "    'bagging_freq': 5,\n",
    "    'min_child_samples': 185,\n",
    "    'min_child_weight': 100.0,\n",
    "    'reg_alpha': 0,\n",
    "    'reg_lambda': 1,\n",
    "    'verbose': 0\n",
    "}\n",
    "\n",
    "gbm_inst = lgb.train(params,\n",
    "                lgb_train,\n",
    "                num_boost_round=20,\n",
    "                valid_sets=lgb_eval)\n",
    "                #early_stopping_rounds=80)\n",
    "\n",
    "pred_inst_gbm = gbm_inst.predict(X_test_inst, num_iteration=gbm_inst.best_iteration)\n",
    "print('The rmse of prediction is:', mean_squared_error(y_test_inst, pred_inst_gbm) ** 0.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "76847.61354285045"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "gbm_inst.save_model('model_light_gbm_inst.txt')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pred_inst_gbm"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Tuneo de hiperparámetros para LightGBM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "gridParams = {  #para el grid search\n",
    "    'learning_rate': [0.05, 0.1, 0.5],\n",
    "    'n_estimators': [1000],\n",
    "    'num_leaves': [5,10,15,20],\n",
    "    'boosting_type' : ['gbdt', 'dart'],\n",
    "    'objective' : ['regression'],\n",
    "    'metric' : ['l2', 'l1'],\n",
    "    'colsample_bytree' : [0.65, 0.66],\n",
    "    'subsample' : [0.3,0.6],\n",
    "    'reg_alpha' : [1,2],\n",
    "    'reg_lambda' : [1,1.5,2],\n",
    "    }\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "param_test ={'num_leaves': sp_randint(6, 50), #para el randomized search\n",
    "             'min_child_samples': sp_randint(100, 500), \n",
    "             'min_child_weight': [1e-5, 1e-3, 1e-2, 1e-1, 1, 1e1, 1e2, 1e3, 1e4],\n",
    "             'reg_alpha': [0, 1e-1, 1, 2, 5, 7, 10, 50, 100],\n",
    "             'reg_lambda': [0, 1e-1, 1, 5, 10, 20, 50, 100]}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> ## Tuneo para el modelo de installs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    ">> #### Randomized Search"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf = lgb.LGBMRegressor(max_depth=-1, random_state=314, silent=True, metric='l2', n_jobs=4, n_estimators = 500)\n",
    "gs_inst = RandomizedSearchCV(\n",
    "    estimator=clf, param_distributions=param_test, \n",
    "    scoring='neg_mean_squared_error',\n",
    "    cv=3,\n",
    "    refit=True,\n",
    "    random_state=314,\n",
    "    verbose=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%time\n",
    "gs_inst.fit(X_train_inst, y_train_inst)\n",
    "print('Best score reached: {} with params: {} '.format(np.sqrt(gs_inst.best_score_*(-1)), gs_inst.best_params_))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    ">> #### Grid Search"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "lgm_grid_inst = lgb.LGBMRegressor(max_depth=-1, learning_rate = 0.01, random_state=314, silent=True, metric='l2', n_jobs=4, n_estimators = 500)\n",
    "\n",
    "gsearchlgm = GridSearchCV(estimator = lgm_grid_inst, param_grid = gridParams,\\\n",
    "                          scoring= 'neg_mean_squared_error', n_jobs=4,iid=False, cv=4)\n",
    "gsearchlgm.fit(df_full_installs[features_list_class_inst], df_full_installs['label_inst'])\n",
    "print('Best score reached: {} with params: {} '.format(np.sqrt(gsearchlgm.best_score_*(-1)), gsearchlgm.best_params_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print('Best score reached: {} with params: {} '.format(np.sqrt(gsearchlgm.best_score_*(-1)), gsearchlgm.best_params_))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> ## Tuneo para el modelo de auctions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    ">> #### Randomized Search"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf_auc = lgb.LGBMRegressor(max_depth=-1, random_state=314, silent=True, metric='l2', n_jobs=4, n_estimators = 500)\n",
    "gs_auc = RandomizedSearchCV(\n",
    "    estimator=clf_auc, param_distributions=param_test, \n",
    "    scoring='neg_mean_squared_error',\n",
    "    cv=3,\n",
    "    refit=True,\n",
    "    random_state=314,\n",
    "    verbose=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_full_auctions = df_full_auctions.sample(160000)\n",
    "X_train_auc, X_test_auc, y_train_auc, y_test_auc = get_train_test_split(df_full_auctions, \"label_auc\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%time\n",
    "gs_auc.fit(X_train_auc, y_train_auc)\n",
    "print('Best score reached: {} with params: {} '.format(np.sqrt(gs_auc.best_score_*(-1)), gs_auc.best_params_))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    ">> #### Grid Search"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_full_auctions = df_full_auctions.sample(100000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "lgm_grid_auc = lgb.LGBMRegressor(max_depth=-1, random_state=314, silent=True, metric='l2', n_jobs=4, n_estimators = 500)\n",
    "\n",
    "gsearchlgm_auc = GridSearchCV(estimator = lgm_grid_auc, param_grid = gridParams,\\\n",
    "                          scoring= 'neg_mean_squared_error', n_jobs=4,iid=False, cv=4)\n",
    "gsearchlgm_auc.fit(df_full_auctions[features_list_class_auc],df_full_auctions['label_auc'])\n",
    "print('Best score reached: {} with params: {} '.format(np.sqrt(gsearchlgm_auc.best_score_*(-1)), gsearchlgm_auc.best_params_))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# **Random Forest**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Random Forest para auctions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_full_auctions = get_df_to_train(windows, features, label_auc, features_list_class_auc)\n",
    "df_full_auctions = df_full_auctions[df_full_auctions[\"label_auc\"] != 259200]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train_auc, X_test_auc, y_train_auc, y_test_auc = get_train_test_split(df_full_auctions, \"label_auc\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "random_forest_auc = RandomForestRegressor(\n",
    "                           n_estimators=100, \n",
    "                           min_samples_split = 10, \n",
    "                           random_state=100\n",
    "                         )\n",
    "\n",
    "#random forest no maneja nulos\n",
    "X_train_auc = X_train_auc.fillna(0)\n",
    "X_test_auc = X_test_auc.fillna(0) \n",
    "\n",
    "random_forest_auc.fit(X_train_auc, y_train_auc.values.ravel())\n",
    "\n",
    "params = {'objective': 'reg:squarederror',\n",
    "          'eta': 0.3,\n",
    "          'max_depth': 5,\n",
    "          'min_child_weight': 3,\n",
    "          'silent': 1,\n",
    "          'subsample': 0.7,\n",
    "          'colsample_bytree': 0.7,\n",
    "          'seed': 1}\n",
    "\n",
    "num_trees=250\n",
    "\n",
    "prediction_auc = random_forest_auc.predict(X_test_auc)\n",
    "\n",
    "rmse = np.sqrt(mean_squared_error(y_test_auc, prediction_auc))\n",
    "print(\"RMSE: %f\" % (rmse))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Random Forest para installs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_full_inst = get_df_to_train(windows, features, label_inst, features_list_class_inst)\n",
    "df_full_inst = df_full_inst[df_full_inst[\"label_inst\"] != 259200]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train_inst, X_test_inst, y_train_inst, y_test_inst = get_train_test_split(df_full_inst, \"label_inst\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "random_forest_inst = RandomForestRegressor(\n",
    "                           n_estimators=1000, \n",
    "                           min_samples_split = 10, \n",
    "                           random_state=100\n",
    "                         )\n",
    "\n",
    "#random forest no maneja nulos\n",
    "X_train_inst = X_train_inst.fillna(0)\n",
    "X_test_inst = X_test_inst.fillna(0) \n",
    "\n",
    "random_forest_inst.fit(X_train_inst, y_train_inst.values.ravel())\n",
    "\n",
    "params = {'objective': 'reg:squarederror',\n",
    "          'eta': 0.3,\n",
    "          'max_depth': 5,\n",
    "          'min_child_weight': 3,\n",
    "          'silent': 1,\n",
    "          'subsample': 0.7,\n",
    "          'colsample_bytree': 0.7,\n",
    "          'seed': 1}\n",
    "\n",
    "num_trees=250\n",
    "\n",
    "prediction_inst = random_forest_inst.predict(X_test_inst)\n",
    "\n",
    "rmse = np.sqrt(mean_squared_error(y_test_inst, prediction_inst))\n",
    "print(\"RMSE: %f\" % (rmse))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Predecir"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Predicción auctions XGBoost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pred_auctions = model_auc.predict(select_features(features_to_predict, features_list_class_auc))\n",
    "df_preds_auctions = pd.DataFrame({'ref_hash' : features_to_predict.index, 'obj' : pred_auctions})\n",
    "df_preds_auctions.to_csv(\"auctions_predictions_xgboost.csv\", index=False) #persistimos los resultados"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Predicción auctions Redes Neuronales"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pred_auctions_neural = model_auc_neural.predict(select_features(features_to_predict, features_list_class_auc).fillna(0))\n",
    "pred_auctions_neural = np.array(pred_auctions_neural).flatten()\n",
    "df_preds_auctions_neural = pd.DataFrame({'ref_hash' : features_to_predict.index, 'obj' : pred_auctions_neural})\n",
    "df_preds_auctions_neural.to_csv(\"auctions_predictions_neural.csv\", index=False) #persistimos los resultados"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Predicción auctions LightGBM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pred_auc_gbm = gbm_auc.predict(select_features(features_to_predict,features_list_class_auc),\\\n",
    "                               num_iteration=gbm_auc.best_iteration)\n",
    "df_preds_auc_gbm = pd.DataFrame({'ref_hash' : features_to_predict.index, 'obj' : pred_auc_gbm})\n",
    "df_preds_auc_gbm.to_csv(\"auctions_predictions_lightgbm.csv\", index=False) #persistimos los resultados"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Predicción auctions Random Forests"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "preds_auc_rf = random_forest_auc.predict(select_features(features_to_predict,features_list_class_auc).fillna(0))\n",
    "df_preds_auc_rf = pd.DataFrame({'ref_hash' : features_to_predict.index, 'obj' : preds_auc_rf})\n",
    "df_preds_auc_rf.to_csv(\"auctions_predictions_randomForest.csv\", index=False) #persistimos los resultados"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Predicción installs XGBoost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pred_installs = model_inst.predict(select_features(features_to_predict, features_list_class_inst))\n",
    "df_preds_installs = pd.DataFrame({'ref_hash' : features_to_predict.index, 'obj' : pred_installs})\n",
    "df_preds_installs.to_csv(\"installs_predictions_xgboost.csv\", index=False) #persistimos los resultados"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Predicción installs Redes Neuronales"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pred_installs_neural = model_inst_neural.predict(select_features(features_to_predict, features_list_class_inst).fillna(0))\n",
    "pred_installs_neural = np.array(pred_installs_neural).flatten()\n",
    "df_preds_installs_neural = pd.DataFrame({'ref_hash' : features_to_predict.index, 'obj' : pred_installs_neural})\n",
    "df_preds_installs_neural.to_csv(\"installs_predictions_neural.csv\", index=False) #persistimos los resultados"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Predicción installs LightGBM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pred_inst_gbm = gbm_inst.predict(select_features(features_to_predict,features_list_class_inst),\\\n",
    "                               num_iteration=gbm_inst.best_iteration)\n",
    "df_preds_inst_gbm = pd.DataFrame({'ref_hash' : features_to_predict.index, 'obj' : pred_inst_gbm})\n",
    "df_preds_inst_gbm.to_csv(\"installs_predictions_lightgbm.csv\", index=False) #persistimos los resultados"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Predicción installs Random Forests"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "preds_inst_rf = random_forest_inst.predict(select_features(features_to_predict,features_list_class_inst).fillna(0))\n",
    "df_preds_inst_rf = pd.DataFrame({'ref_hash' : features_to_predict.index, 'obj' : preds_inst_rf})\n",
    "df_preds_inst_rf.to_csv(\"installs_predictions_randomForest.csv\", index=False) #persistimos los resultados"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Submit to Kaggle"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def export_df(df, name):\n",
    "    df.to_csv(name, index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Las predicciones tendrán seteadas como índice los ref_hash para no perder la referencia\n",
    "No es necesario filtrar los ref_hash y quedarnos solo con los target en las predicciones que obtenemos ya que de eso\n",
    "se encarga la función "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "target = pd.read_csv(\"target_competencia_ids.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#df_preds_installs = pd.read_csv(\"installs_predictions.csv\")\n",
    "#df_preds_auctions = pd.read_csv(\"auctions_predictions.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def create_submit_df(auctions_predictions, installs_predictions, target):\n",
    "    \n",
    "    target = target.set_index('ref_hash')\n",
    "    \n",
    "    auc = auctions_predictions.reset_index()\n",
    "    auc.columns = ['ref_hash','obj']\n",
    "    auc['ref_hash'] = auc['ref_hash'].astype(str) + \"_st\"\n",
    "    auc = auc.set_index('ref_hash')\n",
    "    \n",
    "    ins = installs_predictions.reset_index()\n",
    "    ins.columns = ['ref_hash','obj']\n",
    "    ins['ref_hash'] = ins['ref_hash'].astype(str) + \"_sc\"\n",
    "    ins = ins.set_index('ref_hash')\n",
    "    \n",
    "    frames = [ins,auc]\n",
    "    submit_result = pd.concat(frames).reset_index()\n",
    "    target_list = target.reset_index('ref_hash')['ref_hash'].tolist()\n",
    "    return submit_result.loc[submit_result['ref_hash'].isin(target_list)].sort_values(by = 'ref_hash')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### XGBoost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "kaggle_sub = create_submit_df(df_preds_auctions.set_index('ref_hash'), \\\n",
    "                              df_preds_installs.set_index('ref_hash'), target)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "export_df(kaggle_sub, \"submit.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Neural network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "kaggle_sub_neural = create_submit_df(df_preds_auctions_neural.set_index('ref_hash'), \\\n",
    "                              df_preds_installs_neural.set_index('ref_hash'), target)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "export_df(kaggle_sub_neural, \"submit_neural.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "ensamble_df = kaggle_sub.merge(kaggle_sub_neural, on=\"ref_hash\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "ensamble_df[\"obj\"] = (ensamble_df[\"obj_x\"] + ensamble_df[\"obj_y\"])/2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "ensamble_kaggle = ensamble_df[[\"ref_hash\", \"obj\"]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "export_df(ensamble_kaggle, \"submit_ensamble.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Ensemble XGBoost + LightGBM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "kaggle_sub_lgbm = create_submit_df(df_preds_auc_gbm.set_index('ref_hash'), df_preds_inst_gbm.set_index('ref_hash'), target)\n",
    "kaggle_sub_xgboost = create_submit_df(df_preds_auctions.set_index('ref_hash'), df_preds_installs.set_index('ref_hash'), target)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "xgboost_submit = kaggle_sub_lgbm.merge(kaggle_sub_xgboost, on = 'ref_hash')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "xgboost_submit['obj'] = (xgboost_submit['obj_x'] + xgboost_submit['obj_y'])/2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "xgboost_submit = xgboost_submit[['ref_hash', 'obj']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "export_df(xgboost_submit, \"submit_leo.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Ensamble lgbm + random_forests + xgboost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "a_inst = pd.read_csv(\"installs_predictions_lightgbm.csv\")\n",
    "b_inst = pd.read_csv(\"installs_predictions_randomForest.csv\")\n",
    "c_inst = pd.read_csv(\"installs_predictions_xgboost.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "a_auc = pd.read_csv(\"auctions_predictions_lightgbm.csv\")\n",
    "b_auc = pd.read_csv(\"auctions_predictions_randomForest.csv\")\n",
    "c_auc = pd.read_csv(\"auctions_predictions_xgboost.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "a_b_inst = a_inst.merge(b_inst, on=\"ref_hash\")\n",
    "all_inst = a_b_inst.merge(c_inst, on = \"ref_hash\")\n",
    "all_inst['obj'] = (all_inst['obj_x'] + all_inst['obj_y'] + all_inst['obj'])/3\n",
    "all_inst = all_inst[['ref_hash', 'obj']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "a_b_auc = a_auc.merge(b_auc, on=\"ref_hash\")\n",
    "all_auc = a_b_auc.merge(c_auc, on = \"ref_hash\")\n",
    "all_auc['obj'] = (all_auc['obj_x'] + all_auc['obj_y'] + all_auc['obj'])/3\n",
    "all_auc = all_auc[['ref_hash', 'obj']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "submit = create_submit_df(all_auc.set_index('ref_hash'), all_inst.set_index('ref_hash'), target)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "export_df(submit, \"submit.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Submit con LightGBM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "export_df(create_submit_df(df_preds_auc_gbm.set_index('ref_hash'), df_preds_inst_gbm.set_index('ref_hash'), target), \"submit_prueba.csv\")"
   ]
  },
  {
   "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": "3.7.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}