test3-Copy1.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "from sklearn.metrics import roc_auc_score\n",
    "from sklearn.model_selection import KFold\n",
    "from sklearn.model_selection import StratifiedKFold,cross_val_score\n",
    "from lightgbm import LGBMClassifier\n",
    "from xgboost import XGBClassifier\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "color = sns.color_palette()\n",
    "sns.set_style('darkgrid')\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "from bayes_opt import BayesianOptimization\n",
    "from tqdm import tqdm_notebook as tqdm\n",
    "\n",
    "import gc\n",
    "gc.enable()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Read data and test\n",
      "data test Shapes :  (307511, 122) (48744, 121)\n",
      "all_data Shape :  (356255, 122)\n",
      "['NAME_CONTRACT_TYPE', 'NAME_TYPE_SUITE', 'NAME_INCOME_TYPE', 'NAME_EDUCATION_TYPE', 'NAME_FAMILY_STATUS', 'NAME_HOUSING_TYPE', 'OCCUPATION_TYPE', 'WEEKDAY_APPR_PROCESS_START', 'ORGANIZATION_TYPE', 'FONDKAPREMONT_MODE', 'HOUSETYPE_MODE', 'WALLSMATERIAL_MODE', 'EMERGENCYSTATE_MODE']\n",
      "all_data Shape :  (356255, 255)\n",
      "all_data Shape :  (356255, 261)\n",
      "bureau  buro_bal Shape :  (1716428, 17) (27299925, 3)\n",
      "['STATUS']\n",
      "['CREDIT_ACTIVE', 'CREDIT_CURRENCY', 'CREDIT_TYPE']\n",
      "bb_agg Shape :  (817395, 13)\n",
      "bureau Shape :  (1716428, 52)\n",
      "bureau_agg Shape :  (305811, 71)\n",
      "bureau_agg Shape :  (305811, 107)\n",
      "bureau_agg Shape :  (305811, 145)\n",
      "bureau_agg Shape :  (305811, 153)\n",
      "bureau_agg Shape :  (305811, 161)\n",
      "bureau_agg Shape :  (305811, 169)\n",
      "bureau_agg Shape :  (305811, 177)\n",
      "all_data Shape :  (356255, 437)\n",
      "prev Shape :  (1670214, 37)\n",
      "['NAME_CONTRACT_TYPE', 'WEEKDAY_APPR_PROCESS_START', 'FLAG_LAST_APPL_PER_CONTRACT', 'NAME_CASH_LOAN_PURPOSE', 'NAME_CONTRACT_STATUS', 'NAME_PAYMENT_TYPE', 'CODE_REJECT_REASON', 'NAME_TYPE_SUITE', 'NAME_CLIENT_TYPE', 'NAME_GOODS_CATEGORY', 'NAME_PORTFOLIO', 'NAME_PRODUCT_TYPE', 'CHANNEL_TYPE', 'NAME_SELLER_INDUSTRY', 'NAME_YIELD_GROUP', 'PRODUCT_COMBINATION']\n",
      "prev_agg Shape :  (338857, 212)\n",
      "prev_agg Shape :  (338857, 266)\n",
      "prev_agg Shape :  (338857, 319)\n",
      "all_data Shape :  (356255, 755)\n",
      "pos Shape :  (10001358, 8)\n",
      "['NAME_CONTRACT_STATUS']\n",
      "pos_agg Shape :  (337252, 26)\n",
      "all_data Shape :  (356255, 781)\n",
      "ins Shape :  (13605401, 8)\n",
      "[]\n",
      "ins_agg Shape :  (339587, 35)\n",
      "all_data Shape :  (356255, 816)\n",
      "cc Shape :  (3840312, 23)\n",
      "['NAME_CONTRACT_STATUS']\n",
      "cc_agg Shape :  (103558, 81)\n",
      "all_data Shape :  (356255, 897)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "7"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def one_hot_encoder(df, nan_as_category = True):\n",
    "    original_columns = list(df.columns)\n",
    "    categorical_columns = [col for col in df.columns if df[col].dtype == 'object']\n",
    "    print (categorical_columns)\n",
    "    df = pd.get_dummies(df, columns= categorical_columns, dummy_na= nan_as_category)\n",
    "    new_columns = [c for c in df.columns if c not in original_columns]\n",
    "    return df, new_columns\n",
    "\n",
    "print('Read data and test')\n",
    "data = pd.read_csv('./input/application_train.csv')\n",
    "test = pd.read_csv('./input/application_test.csv')\n",
    "print('data test Shapes : ', data.shape, test.shape)\n",
    "\n",
    "y = data['TARGET']\n",
    "del data['TARGET']\n",
    "gc.collect()\n",
    "all_data = pd.concat((data, test)).reset_index(drop=True)\n",
    "\n",
    "inc_by_org = all_data[['AMT_INCOME_TOTAL', 'ORGANIZATION_TYPE']].groupby('ORGANIZATION_TYPE').median()['AMT_INCOME_TOTAL']\n",
    "all_data['NEW_INC_BY_ORG'] = all_data['ORGANIZATION_TYPE'].map(inc_by_org)\n",
    "print('all_data Shape : ', all_data.shape)\n",
    "\n",
    "for bin_feature in ['CODE_GENDER', 'FLAG_OWN_CAR', 'FLAG_OWN_REALTY']:\n",
    "    all_data[bin_feature], uniques = pd.factorize(all_data[bin_feature])\n",
    "all_data, cat_cols = one_hot_encoder(all_data)\n",
    "print('all_data Shape : ', all_data.shape)\n",
    "# Some simple new features (percentages)\n",
    "#all_data['DAYS_EMPLOYED'].replace(365243, np.nan, inplace= True)\n",
    "n_data = data.shape[0]\n",
    "\n",
    "all_data['INCOME_DAYS_PERC'] = all_data['AMT_INCOME_TOTAL'] / all_data['DAYS_BIRTH']\n",
    "all_data['DAYS_EMPLOYED_PERC'] = all_data['DAYS_EMPLOYED'] / all_data['DAYS_BIRTH']\n",
    "all_data['INCOME_CREDIT_PERC'] = all_data['AMT_INCOME_TOTAL'] / all_data['AMT_CREDIT']\n",
    "all_data['INCOME_PER_PERSON'] = all_data['AMT_INCOME_TOTAL'] / all_data['CNT_FAM_MEMBERS']\n",
    "all_data['ANNUITY_INCOME_PERC'] = all_data['AMT_ANNUITY'] / all_data['AMT_INCOME_TOTAL']\n",
    "all_data['PAYMENT_RATE'] = all_data['AMT_CREDIT'] / all_data['AMT_ANNUITY']\n",
    "print('all_data Shape : ', all_data.shape)\n",
    "del data,test\n",
    "gc.collect()\n",
    "\n",
    "bureau = pd.read_csv('./input/bureau.csv')\n",
    "buro_bal = pd.read_csv('./input/bureau_balance.csv')\n",
    "print('bureau  buro_bal Shape : ', bureau.shape,buro_bal.shape)\n",
    "buro_bal, bb_cat = one_hot_encoder(buro_bal)\n",
    "bureau, bureau_cat = one_hot_encoder(bureau)\n",
    "bb_aggregations = {'MONTHS_BALANCE': ['min', 'max','mean']}\n",
    "for col in bb_cat:\n",
    "    bb_aggregations[col] = ['mean']\n",
    "    bb_agg = buro_bal.groupby('SK_ID_BUREAU').agg(bb_aggregations)\n",
    "bb_agg.columns = pd.Index([e[0] + \"_\" + e[1].upper() for e in bb_agg.columns.tolist()])\n",
    "bb_agg['BB_COUNT'] = buro_bal.groupby('SK_ID_BUREAU').size()\n",
    "print('bb_agg Shape : ', bb_agg.shape)\n",
    "bureau = bureau.join(bb_agg, how='left', on='SK_ID_BUREAU')\n",
    "bureau.drop(columns= 'SK_ID_BUREAU', inplace= True)\n",
    "print('bureau Shape : ', bureau.shape)\n",
    "del buro_bal,bb_agg\n",
    "gc.collect()\n",
    "bureau['PEC_ANNUITY_DEBT'] = bureau['AMT_ANNUITY']/bureau['AMT_CREDIT_SUM_DEBT']\n",
    "num_aggregations = {\n",
    "    'DAYS_CREDIT': ['min', 'max', 'mean', 'var'],\n",
    "    'CREDIT_DAY_OVERDUE': ['max', 'mean'],\n",
    "    'DAYS_CREDIT_ENDDATE': ['min', 'max', 'mean'],\n",
    "    'AMT_CREDIT_MAX_OVERDUE': ['mean'],\n",
    "    'CNT_CREDIT_PROLONG': ['sum','mean'],\n",
    "    'PEC_ANNUITY_DEBT':['min','max','mean'],\n",
    "    'AMT_CREDIT_SUM': ['max','mean', 'sum'],\n",
    "    'AMT_CREDIT_SUM_DEBT': ['max','mean', 'sum'],\n",
    "    'AMT_CREDIT_SUM_OVERDUE': ['mean'],\n",
    "    'AMT_CREDIT_SUM_LIMIT': ['mean', 'sum'],\n",
    "    'DAYS_CREDIT_UPDATE': ['min','max','mean'],\n",
    "    'AMT_ANNUITY': ['max', 'mean'],\n",
    "    'MONTHS_BALANCE_MIN': ['min'],\n",
    "    'MONTHS_BALANCE_MAX': ['max'],\n",
    "    'MONTHS_BALANCE_MEAN': ['mean'],\n",
    "    'BB_COUNT': ['mean', 'sum']\n",
    "}\n",
    "cat_aggregations = {}\n",
    "for cat in bureau_cat: cat_aggregations[cat] = ['mean']\n",
    "for cat in bb_cat: cat_aggregations[cat + \"_MEAN\"] = ['mean']\n",
    "bureau_agg = bureau.groupby('SK_ID_CURR').agg(dict(num_aggregations, **cat_aggregations))\n",
    "bureau_agg.columns = pd.Index(['BURO_' + e[0] + \"_\" + e[1].upper() for e in bureau_agg.columns.tolist()])\n",
    "bureau_agg['BURO_COUNT'] = bureau.groupby('SK_ID_CURR').size()\n",
    "bureau_agg['BURO_MAX_OVERDUE_DEBT'] = bureau_agg['BURO_AMT_CREDIT_MAX_OVERDUE_MEAN']/bureau_agg['BURO_AMT_CREDIT_SUM_DEBT_MEAN']\n",
    "print('bureau_agg Shape : ', bureau_agg.shape)\n",
    "\n",
    "active = bureau[bureau['CREDIT_ACTIVE_Active'] == 1]\n",
    "active_agg = active.groupby('SK_ID_CURR').agg(num_aggregations)\n",
    "active_agg.columns = pd.Index(['ACT_' + e[0] + \"_\" + e[1].upper() for e in active_agg.columns.tolist()])\n",
    "active_agg['ACT_COUNT'] = active.groupby('SK_ID_CURR').size()\n",
    "bureau_agg = bureau_agg.reset_index().join(active_agg, how='left', on='SK_ID_CURR')\n",
    "print('bureau_agg Shape : ', bureau_agg.shape)\n",
    "del active, active_agg\n",
    "gc.collect()\n",
    "\n",
    "closed = bureau[bureau['CREDIT_ACTIVE_Closed'] == 1]\n",
    "closed_agg = closed.groupby('SK_ID_CURR').agg(dict({'DAYS_ENDDATE_FACT': ['min','max','mean']},**num_aggregations))\n",
    "closed_agg.columns = pd.Index(['CLS_' + e[0] + \"_\" + e[1].upper() for e in closed_agg.columns.tolist()])\n",
    "closed_agg['CLS_COUNT'] = closed.groupby('SK_ID_CURR').size()\n",
    "bureau_agg = bureau_agg.join(closed_agg, how='left', on='SK_ID_CURR')\n",
    "print('bureau_agg Shape : ', bureau_agg.shape)\n",
    "del closed, closed_agg\n",
    "gc.collect()\n",
    "\n",
    "for days in [1462,730,365,183]:\n",
    "    stat = bureau[bureau['DAYS_CREDIT']>-days].copy()\n",
    "    agg = {\n",
    "        'DAYS_CREDIT': ['count','mean','skew'],\n",
    "        #'DAYS_ENDDATE_FACT_diff':['max'],\n",
    "        'AMT_CREDIT_MAX_OVERDUE': ['max'],\n",
    "        #'AMT_CREDIT_SUM_DEBT_diff': ['min', 'max', 'mean', 'sum', 'std', 'median', 'skew', 'last', 'first', ss.kurtosis, np.ptp],\n",
    "        'AMT_CREDIT_SUM':['sum'],\n",
    "        'AMT_CREDIT_SUM_DEBT':['sum'],\n",
    "        'AMT_CREDIT_SUM_LIMIT':['mean'],\n",
    "    }\n",
    "\n",
    "    stat_agg = stat.groupby('SK_ID_CURR').agg(agg)\n",
    "    stat_agg.columns = pd.Index([e[0] + \"_\" + e[1].upper() for e in stat_agg.columns.tolist()])\n",
    "    stat_agg['AMT_CREDIT_SUM_DEBT_sum_ratio'] = stat_agg['AMT_CREDIT_SUM_SUM'] / (stat_agg['AMT_CREDIT_SUM_DEBT_SUM'] + 0.01)\n",
    "    stat_agg.columns = [c + '_{}days_ago'.format(days) for c in stat_agg.columns]\n",
    "    bureau_agg = bureau_agg.join(stat_agg, how='left', on='SK_ID_CURR')  \n",
    "    print('bureau_agg Shape : ', bureau_agg.shape)\n",
    "    del stat, stat_agg\n",
    "    gc.collect()  \n",
    "\n",
    "all_data = all_data.merge(bureau_agg, how='left', on='SK_ID_CURR')\n",
    "print('all_data Shape : ', all_data.shape)\n",
    "del bureau_agg\n",
    "gc.collect()\n",
    "\n",
    "prev = pd.read_csv('./input/previous_application.csv')\n",
    "print('prev Shape : ', prev.shape)\n",
    "prev, cat_cols = one_hot_encoder(prev, nan_as_category= True)\n",
    "# Days 365.243 values -> nan\n",
    "prev['DAYS_FIRST_DRAWING'].replace(365243, np.nan, inplace= True)\n",
    "prev['DAYS_FIRST_DUE'].replace(365243, np.nan, inplace= True)\n",
    "prev['DAYS_LAST_DUE_1ST_VERSION'].replace(365243, np.nan, inplace= True)\n",
    "prev['DAYS_LAST_DUE'].replace(365243, np.nan, inplace= True)\n",
    "prev['DAYS_TERMINATION'].replace(365243, np.nan, inplace= True)\n",
    "# Add feature: value ask / value received percentage\n",
    "prev['APP_CREDIT_PERC'] = prev['AMT_APPLICATION'] / prev['AMT_CREDIT']\n",
    "prev['NEW_PA_AMT_UNREPAID'] = prev[prev['DAYS_LAST_DUE_1ST_VERSION'] > 0]\\\n",
    "    ['DAYS_LAST_DUE_1ST_VERSION']/30*(prev[prev['DAYS_LAST_DUE_1ST_VERSION'] > 0]['AMT_ANNUITY'])\n",
    "# Previous applications numeric features\n",
    "num_aggregations = {\n",
    "    'AMT_ANNUITY': ['min','max', 'mean'],\n",
    "    'AMT_APPLICATION': ['min','max', 'mean'],\n",
    "    'AMT_CREDIT': ['min','max', 'mean'],\n",
    "    'APP_CREDIT_PERC': ['min', 'max', 'mean', 'var'],\n",
    "    'AMT_DOWN_PAYMENT': ['min', 'max', 'mean'],\n",
    "    'AMT_GOODS_PRICE': ['min', 'max', 'mean'],\n",
    "    'HOUR_APPR_PROCESS_START': ['min', 'max', 'mean'],\n",
    "    'RATE_DOWN_PAYMENT': ['min', 'max', 'mean'],\n",
    "    'DAYS_DECISION': ['min', 'max', 'mean'],\n",
    "    'SELLERPLACE_AREA': ['min', 'max', 'mean'],\n",
    "    'DAYS_FIRST_DRAWING': ['min', 'max', 'mean'],\n",
    "    'DAYS_FIRST_DUE': ['min', 'max', 'mean'],\n",
    "    'DAYS_LAST_DUE_1ST_VERSION': ['min', 'max', 'mean'],\n",
    "    'DAYS_LAST_DUE': ['min', 'max', 'mean'],\n",
    "    'DAYS_TERMINATION': ['min', 'max', 'mean'],\n",
    "    'CNT_PAYMENT': ['mean', 'sum'],\n",
    "    'NEW_PA_AMT_UNREPAID':['min','max','mean', 'sum']\n",
    "}\n",
    "# Previous applications categorical features\n",
    "cat_aggregations = {}\n",
    "for cat in cat_cols:\n",
    "    cat_aggregations[cat] = ['mean']\n",
    "\n",
    "prev_agg = prev.groupby('SK_ID_CURR').agg(dict(num_aggregations, **cat_aggregations))\n",
    "prev_agg.columns = pd.Index(['PREV_' + e[0] + \"_\" + e[1].upper() for e in prev_agg.columns.tolist()])\n",
    "prev_agg['PREV_COUNT'] = prev.groupby('SK_ID_CURR').size()\n",
    "print('prev_agg Shape : ', prev_agg.shape)\n",
    "# Previous Applications: Approved Applications - only numerical features\n",
    "\n",
    "ap_num_aggregations = {\n",
    "        'AMT_ANNUITY': ['min','max', 'mean'],\n",
    "        'AMT_APPLICATION': ['min','max', 'mean'],\n",
    "        'AMT_CREDIT': ['min','max', 'mean'],\n",
    "        'APP_CREDIT_PERC': ['min', 'max', 'mean', 'var'],\n",
    "        'AMT_DOWN_PAYMENT': ['min', 'max', 'mean'],\n",
    "        'AMT_GOODS_PRICE': ['min', 'max', 'mean'],\n",
    "        'HOUR_APPR_PROCESS_START': ['min', 'max', 'mean'],\n",
    "        'RATE_DOWN_PAYMENT': ['min', 'max', 'mean'],\n",
    "        'DAYS_DECISION': ['min', 'max', 'mean'],\n",
    "        'SELLERPLACE_AREA': ['min', 'max', 'mean'],\n",
    "        'DAYS_FIRST_DRAWING': ['min', 'max', 'mean'],\n",
    "        'DAYS_FIRST_DUE': ['min', 'max', 'mean'],\n",
    "        'DAYS_LAST_DUE_1ST_VERSION': ['min', 'max', 'mean'],\n",
    "        'DAYS_LAST_DUE': ['min', 'max', 'mean'],\n",
    "        'DAYS_TERMINATION': ['min', 'max', 'mean'],\n",
    "        'CNT_PAYMENT': ['mean', 'sum']\n",
    "    }\n",
    "approved = prev[prev['NAME_CONTRACT_STATUS_Approved'] == 1]\n",
    "approved_agg = approved.groupby('SK_ID_CURR').agg(num_aggregations)\n",
    "approved_agg.columns = pd.Index(['APR_' + e[0] + \"_\" + e[1].upper() for e in approved_agg.columns.tolist()])\n",
    "approved_agg['APR_COUNT'] = approved.groupby('SK_ID_CURR').size()\n",
    "prev_agg = prev_agg.reset_index().join(approved_agg, how='left', on='SK_ID_CURR')\n",
    "print('prev_agg Shape : ', prev_agg.shape)\n",
    "# Previous Applications: Refused Applications - only numerical features\n",
    "refused = prev[prev['NAME_CONTRACT_STATUS_Refused'] == 1]\n",
    "refused_agg = refused.groupby('SK_ID_CURR').agg(num_aggregations)\n",
    "refused_agg.columns = pd.Index(['REF_' + e[0] + \"_\" + e[1].upper() for e in refused_agg.columns.tolist()])\n",
    "refused_agg['REF_COUNT'] = refused.groupby('SK_ID_CURR').size()\n",
    "prev_agg = prev_agg.join(refused_agg, how='left', on='SK_ID_CURR')\n",
    "print('prev_agg Shape : ', prev_agg.shape)\n",
    "del refused, refused_agg, approved, approved_agg, prev\n",
    "gc.collect()\n",
    "\n",
    "all_data = all_data.merge(prev_agg, how='left', on='SK_ID_CURR')\n",
    "print('all_data Shape : ', all_data.shape)\n",
    "del prev_agg\n",
    "gc.collect()\n",
    "\n",
    "pos = pd.read_csv('./input/POS_CASH_balance.csv')\n",
    "print('pos Shape : ', pos.shape)\n",
    "pos, cat_cols = one_hot_encoder(pos, nan_as_category= True)\n",
    "# Features\n",
    "aggregations = {\n",
    "    'MONTHS_BALANCE': ['min','max', 'mean'],\n",
    "    'CNT_INSTALMENT': ['sum', 'mean'],\n",
    "    'SK_DPD': ['sum','max','min','mean','size'],\n",
    "    'SK_DPD_DEF': ['sum','max', 'min','mean','size']\n",
    "}\n",
    "for cat in cat_cols:\n",
    "    aggregations[cat] = ['mean']\n",
    "\n",
    "pos_agg = pos.groupby('SK_ID_CURR').agg(aggregations)\n",
    "pos_agg.columns = pd.Index(['POS_' + e[0] + \"_\" + e[1].upper() for e in pos_agg.columns.tolist()])\n",
    "# Count pos cash accounts\n",
    "pos_agg['POS_COUNT'] = pos.groupby('SK_ID_CURR').size()\n",
    "print('pos_agg Shape : ', pos_agg.shape)\n",
    "del pos\n",
    "gc.collect()\n",
    "\n",
    "all_data = all_data.join(pos_agg, how='left', on='SK_ID_CURR')\n",
    "print('all_data Shape : ', all_data.shape)\n",
    "del pos_agg\n",
    "gc.collect()\n",
    "\n",
    "ins = pd.read_csv('./input/installments_payments.csv')\n",
    "print('ins Shape : ', ins.shape)\n",
    "ins, cat_cols = one_hot_encoder(ins, nan_as_category= True)\n",
    "# Percentage and difference paid in each installment (amount paid and installment value)\n",
    "ins['PAYMENT_PERC'] = ins['AMT_PAYMENT'] / ins['AMT_INSTALMENT']\n",
    "ins['PAYMENT_DIFF'] = ins['AMT_INSTALMENT'] - ins['AMT_PAYMENT']\n",
    "# Days past due and days before due (no negative values)\n",
    "ins['DPD'] = ins['DAYS_ENTRY_PAYMENT'] - ins['DAYS_INSTALMENT']\n",
    "ins['NEW_IP_DPD_RATIO'] = ins['DAYS_ENTRY_PAYMENT'] / ins['DAYS_INSTALMENT']\n",
    "ins['DBD'] = ins['DAYS_INSTALMENT'] - ins['DAYS_ENTRY_PAYMENT']\n",
    "ins['NEW_IP_DBD_RATIO'] = ins['DAYS_INSTALMENT'] / ins['DAYS_ENTRY_PAYMENT']\n",
    "ins['DPD'] = ins['DPD'].apply(lambda x: x if x > 0 else 0)\n",
    "ins['DBD'] = ins['DBD'].apply(lambda x: x if x > 0 else 0)\n",
    "# Features: Perform aggregations\n",
    "aggregations = {\n",
    "    'NUM_INSTALMENT_VERSION': ['nunique'],\n",
    "    'DPD': ['max','mean', 'sum'],\n",
    "    'NEW_IP_DPD_RATIO':['max', 'mean','min'],\n",
    "    'NEW_IP_DBD_RATIO':['max', 'mean','min'],\n",
    "    'DBD': ['max', 'mean', 'sum'],\n",
    "    'PAYMENT_PERC': ['max', 'mean', 'sum', 'var'],\n",
    "    'PAYMENT_DIFF': ['max', 'mean', 'sum', 'var'],\n",
    "    'AMT_INSTALMENT': ['max', 'mean', 'sum'],\n",
    "    'AMT_PAYMENT': ['min', 'max', 'mean', 'sum'],\n",
    "    'DAYS_ENTRY_PAYMENT': ['max', 'mean', 'sum'],\n",
    "    'DAYS_INSTALMENT': ['max', 'mean', 'sum']\n",
    "}\n",
    "for cat in cat_cols:\n",
    "    aggregations[cat] = ['mean']\n",
    "ins_agg = ins.groupby('SK_ID_CURR').agg(aggregations)\n",
    "ins_agg.columns = pd.Index(['INS_' + e[0] + \"_\" + e[1].upper() for e in ins_agg.columns.tolist()])\n",
    "# Count installments accounts\n",
    "ins_agg['INS_COUNT'] = ins.groupby('SK_ID_CURR').size()\n",
    "print('ins_agg Shape : ', ins_agg.shape)\n",
    "del ins\n",
    "gc.collect()\n",
    "\n",
    "all_data = all_data.join(ins_agg, how='left', on='SK_ID_CURR')\n",
    "print('all_data Shape : ', all_data.shape)\n",
    "del ins_agg\n",
    "gc.collect()\n",
    "\n",
    "cc = pd.read_csv('./input/credit_card_balance.csv')\n",
    "print('cc Shape : ', cc.shape)\n",
    "cc, cat_cols = one_hot_encoder(cc, nan_as_category= True)\n",
    "# Features: Perform aggregations\n",
    "num_aggregations = {\n",
    "    'MONTHS_BALANCE': ['min','max', 'mean'],\n",
    "    'AMT_BALANCE': ['min','max', 'mean','sum'],\n",
    "    'AMT_CREDIT_LIMIT_ACTUAL': ['min','max', 'mean','sum','var'],\n",
    "    'AMT_DRAWINGS_ATM_CURRENT': ['min','max', 'mean','sum'],\n",
    "    'AMT_DRAWINGS_CURRENT': ['min','max', 'mean','sum'],\n",
    "    'AMT_DRAWINGS_OTHER_CURRENT': ['min','max', 'mean','sum'],\n",
    "    'AMT_DRAWINGS_POS_CURRENT': ['min','max', 'mean','sum'],\n",
    "    'AMT_INST_MIN_REGULARITY': ['min','max', 'mean','sum','var'],\n",
    "    'AMT_PAYMENT_CURRENT':['min','max', 'mean','sum'],\n",
    "    'AMT_PAYMENT_TOTAL_CURRENT':['min','max', 'mean','sum'],\n",
    "    'AMT_RECEIVABLE_PRINCIPAL':['min','max', 'mean','sum','var'],\n",
    "    'AMT_RECIVABLE':['min','max', 'mean','sum','var'],\n",
    "    'AMT_TOTAL_RECEIVABLE':['min','max', 'mean','sum','var'],\n",
    "    'CNT_DRAWINGS_ATM_CURRENT':['mean','sum'],\n",
    "    'CNT_DRAWINGS_CURRENT':['mean','sum'],\n",
    "    'CNT_DRAWINGS_OTHER_CURRENT':['mean','sum'],\n",
    "    'CNT_DRAWINGS_POS_CURRENT':['mean','sum'],\n",
    "    'CNT_INSTALMENT_MATURE_CUM':['mean','sum'],\n",
    "    'SK_DPD':['max', 'mean', 'sum'],\n",
    "    'SK_DPD_DEF':['max', 'mean', 'sum']\n",
    "}\n",
    "cat_aggregations = {}\n",
    "for cat in cat_cols:\n",
    "    cat_aggregations[cat] = ['mean']\n",
    "# General aggregations\n",
    "cc_agg = cc.groupby('SK_ID_CURR').agg(dict(num_aggregations,**cat_aggregations))\n",
    "cc_agg.columns = pd.Index(['CC_' + e[0] + \"_\" + e[1].upper() for e in cc_agg.columns.tolist()])\n",
    "# Count credit card lines\n",
    "cc_agg['CC_COUNT'] = cc.groupby('SK_ID_CURR').size()\n",
    "\n",
    "#cc_agg['INSTALLMENTS_PER_LOAN'] = cc_agg['CNT_INSTALMENT_MATURE_CUM_SUM'] / cc_agg['CC_COUNT']\n",
    "del cc\n",
    "print('cc_agg Shape : ', cc_agg.shape)\n",
    "gc.collect()\n",
    "\n",
    "all_data = all_data.join(cc_agg, how='left', on='SK_ID_CURR')\n",
    "print('all_data Shape : ', all_data.shape)\n",
    "del cc_agg\n",
    "gc.collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(307511, 879)\n",
      "(48744, 879)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "['SK_ID_CURR']"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "all_data_na = (all_data.isnull().sum() / len(all_data)) * 100\n",
    "all_data_na = all_data_na.drop(all_data_na[all_data_na == 0].index).sort_values(ascending=False)\n",
    "missing_data = pd.DataFrame({'Missing Ratio' :all_data_na})\n",
    "\n",
    "missing_data\n",
    "\n",
    "drop_col = missing_data[missing_data['Missing Ratio']==100]\n",
    "missing_data = missing_data[missing_data['Missing Ratio']<100]\n",
    "\n",
    "all_data.drop(drop_col.index.values,1,inplace=True)\n",
    "\n",
    "data = all_data[:n_data]\n",
    "test = all_data[n_data:]\n",
    "#del avg_buro, avg_prev,avg_pos,avg_cc_bal,avg_inst\n",
    "del all_data\n",
    "gc.collect()\n",
    "print (data.shape)\n",
    "print (test.shape)\n",
    "\n",
    "exclude_fea = []\n",
    "exclude_fea.append('SK_ID_CURR') \n",
    "exclude_fea"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "folds = KFold(n_splits=5, shuffle=True, random_state=1700)\n",
    "oof_preds = np.zeros(data.shape[0])\n",
    "sub_preds = np.zeros(test.shape[0])\n",
    "feature_importance_df = pd.DataFrame()\n",
    "feats = [f for f in data.columns if f not in exclude_fea]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "NFOLDS=5\n",
    "SEED = 1700\n",
    "folds = KFold(n_splits=NFOLDS, shuffle=True, random_state=SEED)\n",
    "feats = [f for f in data.columns if f not in exclude_fea]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "def get_lgb_oof(x_train, y_train, x_test):\n",
    "    oof_train = np.zeros((x_train.shape[0],))\n",
    "    oof_test = np.zeros((x_test.shape[0],))\n",
    "    oof_test_skf = np.empty((NFOLDS, x_test.shape[0]))\n",
    "\n",
    "    for i, (train_index, test_index) in enumerate(folds.split(x_train)):\n",
    "        x_tr = x_train[feats].iloc[train_index]\n",
    "        y_tr = y_train.iloc[train_index]\n",
    "        x_te = x_train[feats].iloc[test_index]\n",
    "        y_te = y_train.iloc[test_index]\n",
    "        clf = LGBMClassifier(\n",
    "            n_estimators=10000,\n",
    "            learning_rate=0.01,\n",
    "            num_leaves=31,\n",
    "            colsample_bytree=0.9338690,\n",
    "            subsample=0.97291513,\n",
    "            max_depth=8,\n",
    "            reg_alpha=0.06663684,\n",
    "            reg_lambda=0.0988343,\n",
    "            min_split_gain=0.07916152,\n",
    "            min_child_weight=82.8850799,\n",
    "            random_state=1700,\n",
    "        )\n",
    "        clf.fit(x_tr, y_tr, \n",
    "            eval_set= [(x_tr, y_tr), (x_te, y_te)], \n",
    "            eval_metric='auc', verbose=100, early_stopping_rounds=300  #30\n",
    "           )\n",
    "\n",
    "        oof_train[test_index] = clf.predict_proba(x_te, num_iteration=clf.best_iteration_)[:,1]\n",
    "        oof_test_skf[i, :] = clf.predict_proba(x_test[feats], num_iteration=clf.best_iteration_)[:,1]\n",
    "        print('Fold %2d Train AUC : %.6f' % (i + 1, roc_auc_score(y_tr, clf.predict_proba(x_tr, num_iteration=clf.best_iteration_)[:,1])))\n",
    "        print('Fold %2d Test AUC : %.6f' % (i + 1, roc_auc_score(y_te, oof_train[test_index])))\n",
    "        del clf,x_tr,y_tr,x_te,y_te\n",
    "        gc.collect()\n",
    "    oof_test[:] = oof_test_skf.mean(axis=0)\n",
    "    return oof_train.reshape(-1, 1), oof_test.reshape(-1, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training until validation scores don't improve for 300 rounds.\n",
      "[100]\ttraining's auc: 0.742489\tvalid_1's auc: 0.731959\n",
      "[200]\ttraining's auc: 0.75319\tvalid_1's auc: 0.740422\n",
      "[300]\ttraining's auc: 0.768026\tvalid_1's auc: 0.752129\n",
      "[400]\ttraining's auc: 0.783229\tvalid_1's auc: 0.763588\n",
      "[500]\ttraining's auc: 0.794566\tvalid_1's auc: 0.771642\n",
      "[600]\ttraining's auc: 0.802785\tvalid_1's auc: 0.777209\n",
      "[700]\ttraining's auc: 0.809425\tvalid_1's auc: 0.78123\n",
      "[800]\ttraining's auc: 0.814982\tvalid_1's auc: 0.784034\n",
      "[900]\ttraining's auc: 0.819639\tvalid_1's auc: 0.786062\n",
      "[1000]\ttraining's auc: 0.823834\tvalid_1's auc: 0.7876\n",
      "[1100]\ttraining's auc: 0.827775\tvalid_1's auc: 0.788714\n",
      "[1200]\ttraining's auc: 0.831249\tvalid_1's auc: 0.789611\n",
      "[1300]\ttraining's auc: 0.834535\tvalid_1's auc: 0.790387\n",
      "[1400]\ttraining's auc: 0.837581\tvalid_1's auc: 0.791104\n",
      "[1500]\ttraining's auc: 0.840355\tvalid_1's auc: 0.791689\n",
      "[1600]\ttraining's auc: 0.843004\tvalid_1's auc: 0.792272\n",
      "[1700]\ttraining's auc: 0.845692\tvalid_1's auc: 0.792755\n",
      "[1800]\ttraining's auc: 0.848216\tvalid_1's auc: 0.793171\n",
      "[1900]\ttraining's auc: 0.850682\tvalid_1's auc: 0.793513\n",
      "[2000]\ttraining's auc: 0.852942\tvalid_1's auc: 0.793844\n",
      "[2100]\ttraining's auc: 0.855213\tvalid_1's auc: 0.794205\n",
      "[2200]\ttraining's auc: 0.857554\tvalid_1's auc: 0.794491\n",
      "[2300]\ttraining's auc: 0.859889\tvalid_1's auc: 0.794659\n",
      "[2400]\ttraining's auc: 0.861913\tvalid_1's auc: 0.794842\n",
      "[2500]\ttraining's auc: 0.863852\tvalid_1's auc: 0.794992\n",
      "[2600]\ttraining's auc: 0.865885\tvalid_1's auc: 0.795102\n",
      "[2700]\ttraining's auc: 0.867918\tvalid_1's auc: 0.795178\n",
      "[2800]\ttraining's auc: 0.86979\tvalid_1's auc: 0.795226\n",
      "[2900]\ttraining's auc: 0.871585\tvalid_1's auc: 0.795342\n",
      "[3000]\ttraining's auc: 0.87333\tvalid_1's auc: 0.795473\n",
      "[3100]\ttraining's auc: 0.875099\tvalid_1's auc: 0.795591\n",
      "[3200]\ttraining's auc: 0.876883\tvalid_1's auc: 0.795705\n",
      "[3300]\ttraining's auc: 0.878564\tvalid_1's auc: 0.7958\n",
      "[3400]\ttraining's auc: 0.880356\tvalid_1's auc: 0.795876\n",
      "[3500]\ttraining's auc: 0.88198\tvalid_1's auc: 0.795991\n",
      "[3600]\ttraining's auc: 0.88378\tvalid_1's auc: 0.796065\n",
      "[3700]\ttraining's auc: 0.885397\tvalid_1's auc: 0.796128\n",
      "[3800]\ttraining's auc: 0.886942\tvalid_1's auc: 0.796141\n",
      "[3900]\ttraining's auc: 0.88865\tvalid_1's auc: 0.796181\n",
      "[4000]\ttraining's auc: 0.890221\tvalid_1's auc: 0.796242\n",
      "[4100]\ttraining's auc: 0.891875\tvalid_1's auc: 0.796348\n",
      "[4200]\ttraining's auc: 0.89328\tvalid_1's auc: 0.796369\n",
      "[4300]\ttraining's auc: 0.894768\tvalid_1's auc: 0.796361\n",
      "[4400]\ttraining's auc: 0.896323\tvalid_1's auc: 0.796359\n",
      "[4500]\ttraining's auc: 0.897767\tvalid_1's auc: 0.796303\n",
      "Early stopping, best iteration is:\n",
      "[4281]\ttraining's auc: 0.894475\tvalid_1's auc: 0.796379\n",
      "Fold  1 Train AUC : 0.894475\n",
      "Fold  1 Test AUC : 0.796380\n",
      "Training until validation scores don't improve for 300 rounds.\n",
      "[100]\ttraining's auc: 0.742085\tvalid_1's auc: 0.737117\n",
      "[200]\ttraining's auc: 0.752069\tvalid_1's auc: 0.744593\n",
      "[300]\ttraining's auc: 0.766222\tvalid_1's auc: 0.756181\n",
      "[400]\ttraining's auc: 0.781932\tvalid_1's auc: 0.769613\n",
      "[500]\ttraining's auc: 0.793209\tvalid_1's auc: 0.777986\n",
      "[600]\ttraining's auc: 0.801729\tvalid_1's auc: 0.78361\n",
      "[700]\ttraining's auc: 0.808454\tvalid_1's auc: 0.787283\n",
      "[800]\ttraining's auc: 0.814035\tvalid_1's auc: 0.789776\n",
      "[900]\ttraining's auc: 0.818704\tvalid_1's auc: 0.79143\n",
      "[1000]\ttraining's auc: 0.822883\tvalid_1's auc: 0.792706\n",
      "[1100]\ttraining's auc: 0.826536\tvalid_1's auc: 0.793685\n",
      "[1200]\ttraining's auc: 0.829936\tvalid_1's auc: 0.794584\n",
      "[1300]\ttraining's auc: 0.833156\tvalid_1's auc: 0.795263\n",
      "[1400]\ttraining's auc: 0.836153\tvalid_1's auc: 0.795904\n",
      "[1500]\ttraining's auc: 0.839212\tvalid_1's auc: 0.796482\n",
      "[1600]\ttraining's auc: 0.841876\tvalid_1's auc: 0.796847\n",
      "[1700]\ttraining's auc: 0.844605\tvalid_1's auc: 0.797315\n",
      "[1800]\ttraining's auc: 0.847073\tvalid_1's auc: 0.797628\n",
      "[1900]\ttraining's auc: 0.849649\tvalid_1's auc: 0.797976\n",
      "[2000]\ttraining's auc: 0.852138\tvalid_1's auc: 0.798243\n",
      "[2100]\ttraining's auc: 0.854448\tvalid_1's auc: 0.798474\n",
      "[2200]\ttraining's auc: 0.856693\tvalid_1's auc: 0.798662\n",
      "[2300]\ttraining's auc: 0.858855\tvalid_1's auc: 0.798733\n",
      "[2400]\ttraining's auc: 0.860996\tvalid_1's auc: 0.798955\n",
      "[2500]\ttraining's auc: 0.863133\tvalid_1's auc: 0.799154\n",
      "[2600]\ttraining's auc: 0.864935\tvalid_1's auc: 0.799292\n",
      "[2700]\ttraining's auc: 0.866871\tvalid_1's auc: 0.799383\n",
      "[2800]\ttraining's auc: 0.868882\tvalid_1's auc: 0.799525\n",
      "[2900]\ttraining's auc: 0.870794\tvalid_1's auc: 0.799622\n",
      "[3000]\ttraining's auc: 0.872675\tvalid_1's auc: 0.799635\n",
      "[3100]\ttraining's auc: 0.874482\tvalid_1's auc: 0.799694\n",
      "[3200]\ttraining's auc: 0.876292\tvalid_1's auc: 0.799713\n",
      "[3300]\ttraining's auc: 0.878065\tvalid_1's auc: 0.799779\n",
      "[3400]\ttraining's auc: 0.879867\tvalid_1's auc: 0.799821\n",
      "[3500]\ttraining's auc: 0.88163\tvalid_1's auc: 0.800002\n",
      "[3600]\ttraining's auc: 0.883363\tvalid_1's auc: 0.800031\n",
      "[3700]\ttraining's auc: 0.885024\tvalid_1's auc: 0.800044\n",
      "[3800]\ttraining's auc: 0.886675\tvalid_1's auc: 0.800116\n",
      "[3900]\ttraining's auc: 0.888364\tvalid_1's auc: 0.800086\n",
      "[4000]\ttraining's auc: 0.889884\tvalid_1's auc: 0.800094\n",
      "[4100]\ttraining's auc: 0.891425\tvalid_1's auc: 0.800149\n",
      "[4200]\ttraining's auc: 0.892899\tvalid_1's auc: 0.800118\n",
      "[4300]\ttraining's auc: 0.894324\tvalid_1's auc: 0.800141\n",
      "[4400]\ttraining's auc: 0.895853\tvalid_1's auc: 0.800139\n",
      "Early stopping, best iteration is:\n",
      "[4114]\ttraining's auc: 0.891635\tvalid_1's auc: 0.800163\n",
      "Fold  2 Train AUC : 0.891634\n",
      "Fold  2 Test AUC : 0.800161\n",
      "Training until validation scores don't improve for 300 rounds.\n",
      "[100]\ttraining's auc: 0.742028\tvalid_1's auc: 0.736963\n",
      "[200]\ttraining's auc: 0.752974\tvalid_1's auc: 0.745035\n",
      "[300]\ttraining's auc: 0.767694\tvalid_1's auc: 0.755735\n",
      "[400]\ttraining's auc: 0.78313\tvalid_1's auc: 0.767433\n",
      "[500]\ttraining's auc: 0.794508\tvalid_1's auc: 0.774966\n",
      "[600]\ttraining's auc: 0.803007\tvalid_1's auc: 0.779895\n",
      "[700]\ttraining's auc: 0.80971\tvalid_1's auc: 0.783034\n",
      "[800]\ttraining's auc: 0.815199\tvalid_1's auc: 0.785262\n",
      "[900]\ttraining's auc: 0.820005\tvalid_1's auc: 0.786993\n",
      "[1000]\ttraining's auc: 0.824235\tvalid_1's auc: 0.788295\n",
      "[1100]\ttraining's auc: 0.827929\tvalid_1's auc: 0.789162\n",
      "[1200]\ttraining's auc: 0.831396\tvalid_1's auc: 0.789934\n",
      "[1300]\ttraining's auc: 0.834567\tvalid_1's auc: 0.790541\n",
      "[1400]\ttraining's auc: 0.837522\tvalid_1's auc: 0.790975\n",
      "[1500]\ttraining's auc: 0.840337\tvalid_1's auc: 0.791423\n",
      "[1600]\ttraining's auc: 0.843045\tvalid_1's auc: 0.791829\n",
      "[1700]\ttraining's auc: 0.845563\tvalid_1's auc: 0.792103\n",
      "[1800]\ttraining's auc: 0.848045\tvalid_1's auc: 0.792408\n",
      "[1900]\ttraining's auc: 0.850552\tvalid_1's auc: 0.792595\n",
      "[2000]\ttraining's auc: 0.852886\tvalid_1's auc: 0.792793\n",
      "[2100]\ttraining's auc: 0.855233\tvalid_1's auc: 0.793013\n",
      "[2200]\ttraining's auc: 0.857303\tvalid_1's auc: 0.793209\n",
      "[2300]\ttraining's auc: 0.859492\tvalid_1's auc: 0.793428\n",
      "[2400]\ttraining's auc: 0.861601\tvalid_1's auc: 0.793632\n",
      "[2500]\ttraining's auc: 0.863701\tvalid_1's auc: 0.793768\n",
      "[2600]\ttraining's auc: 0.865788\tvalid_1's auc: 0.793784\n",
      "[2700]\ttraining's auc: 0.867732\tvalid_1's auc: 0.793845\n",
      "[2800]\ttraining's auc: 0.869709\tvalid_1's auc: 0.79391\n",
      "[2900]\ttraining's auc: 0.87166\tvalid_1's auc: 0.793963\n",
      "[3000]\ttraining's auc: 0.873607\tvalid_1's auc: 0.794044\n",
      "[3100]\ttraining's auc: 0.875549\tvalid_1's auc: 0.794079\n",
      "[3200]\ttraining's auc: 0.877352\tvalid_1's auc: 0.794055\n",
      "[3300]\ttraining's auc: 0.879085\tvalid_1's auc: 0.794138\n",
      "[3400]\ttraining's auc: 0.880915\tvalid_1's auc: 0.79416\n",
      "[3500]\ttraining's auc: 0.88259\tvalid_1's auc: 0.794181\n",
      "[3600]\ttraining's auc: 0.884236\tvalid_1's auc: 0.794157\n",
      "[3700]\ttraining's auc: 0.885927\tvalid_1's auc: 0.794186\n",
      "[3800]\ttraining's auc: 0.887563\tvalid_1's auc: 0.794166\n",
      "[3900]\ttraining's auc: 0.889194\tvalid_1's auc: 0.794171\n",
      "Early stopping, best iteration is:\n",
      "[3665]\ttraining's auc: 0.88529\tvalid_1's auc: 0.794221\n",
      "Fold  3 Train AUC : 0.885289\n",
      "Fold  3 Test AUC : 0.794222\n",
      "Training until validation scores don't improve for 300 rounds.\n",
      "[100]\ttraining's auc: 0.74185\tvalid_1's auc: 0.733395\n",
      "[200]\ttraining's auc: 0.75306\tvalid_1's auc: 0.74267\n",
      "[300]\ttraining's auc: 0.767215\tvalid_1's auc: 0.754295\n",
      "[400]\ttraining's auc: 0.782856\tvalid_1's auc: 0.767125\n",
      "[500]\ttraining's auc: 0.794391\tvalid_1's auc: 0.775201\n",
      "[600]\ttraining's auc: 0.802784\tvalid_1's auc: 0.780351\n",
      "[700]\ttraining's auc: 0.809451\tvalid_1's auc: 0.783767\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[800]\ttraining's auc: 0.814914\tvalid_1's auc: 0.786158\n",
      "[900]\ttraining's auc: 0.81965\tvalid_1's auc: 0.78815\n",
      "[1000]\ttraining's auc: 0.823795\tvalid_1's auc: 0.789584\n",
      "[1100]\ttraining's auc: 0.827507\tvalid_1's auc: 0.79059\n",
      "[1200]\ttraining's auc: 0.830976\tvalid_1's auc: 0.791381\n",
      "[1300]\ttraining's auc: 0.834139\tvalid_1's auc: 0.792046\n",
      "[1400]\ttraining's auc: 0.837076\tvalid_1's auc: 0.792568\n",
      "[1500]\ttraining's auc: 0.840058\tvalid_1's auc: 0.792977\n",
      "[1600]\ttraining's auc: 0.842818\tvalid_1's auc: 0.793417\n",
      "[1700]\ttraining's auc: 0.845466\tvalid_1's auc: 0.793789\n",
      "[1800]\ttraining's auc: 0.848097\tvalid_1's auc: 0.794027\n",
      "[1900]\ttraining's auc: 0.850521\tvalid_1's auc: 0.794258\n",
      "[2000]\ttraining's auc: 0.85298\tvalid_1's auc: 0.794532\n",
      "[2100]\ttraining's auc: 0.855424\tvalid_1's auc: 0.794817\n",
      "[2200]\ttraining's auc: 0.85773\tvalid_1's auc: 0.795008\n",
      "[2300]\ttraining's auc: 0.859903\tvalid_1's auc: 0.795182\n",
      "[2400]\ttraining's auc: 0.862001\tvalid_1's auc: 0.795265\n",
      "[2500]\ttraining's auc: 0.864057\tvalid_1's auc: 0.795357\n",
      "[2600]\ttraining's auc: 0.866084\tvalid_1's auc: 0.795458\n",
      "[2700]\ttraining's auc: 0.868055\tvalid_1's auc: 0.795513\n",
      "[2800]\ttraining's auc: 0.869992\tvalid_1's auc: 0.795617\n",
      "[2900]\ttraining's auc: 0.871999\tvalid_1's auc: 0.795622\n",
      "[3000]\ttraining's auc: 0.873822\tvalid_1's auc: 0.795662\n",
      "[3100]\ttraining's auc: 0.875648\tvalid_1's auc: 0.795637\n",
      "[3200]\ttraining's auc: 0.877457\tvalid_1's auc: 0.795662\n",
      "[3300]\ttraining's auc: 0.879237\tvalid_1's auc: 0.795721\n",
      "[3400]\ttraining's auc: 0.881054\tvalid_1's auc: 0.795778\n",
      "[3500]\ttraining's auc: 0.882838\tvalid_1's auc: 0.795793\n",
      "[3600]\ttraining's auc: 0.884581\tvalid_1's auc: 0.795798\n",
      "[3700]\ttraining's auc: 0.886233\tvalid_1's auc: 0.795789\n",
      "[3800]\ttraining's auc: 0.888054\tvalid_1's auc: 0.795715\n",
      "Early stopping, best iteration is:\n",
      "[3560]\ttraining's auc: 0.883802\tvalid_1's auc: 0.795852\n",
      "Fold  4 Train AUC : 0.883799\n",
      "Fold  4 Test AUC : 0.795849\n",
      "Training until validation scores don't improve for 300 rounds.\n",
      "[100]\ttraining's auc: 0.744616\tvalid_1's auc: 0.730094\n",
      "[200]\ttraining's auc: 0.755748\tvalid_1's auc: 0.739272\n",
      "[300]\ttraining's auc: 0.769298\tvalid_1's auc: 0.749337\n",
      "[400]\ttraining's auc: 0.784263\tvalid_1's auc: 0.760283\n",
      "[500]\ttraining's auc: 0.795539\tvalid_1's auc: 0.76824\n",
      "[600]\ttraining's auc: 0.803864\tvalid_1's auc: 0.773619\n",
      "[700]\ttraining's auc: 0.810334\tvalid_1's auc: 0.777448\n",
      "[800]\ttraining's auc: 0.815768\tvalid_1's auc: 0.780159\n",
      "[900]\ttraining's auc: 0.820328\tvalid_1's auc: 0.782246\n",
      "[1000]\ttraining's auc: 0.824499\tvalid_1's auc: 0.783839\n",
      "[1100]\ttraining's auc: 0.828198\tvalid_1's auc: 0.784991\n",
      "[1200]\ttraining's auc: 0.831726\tvalid_1's auc: 0.785926\n",
      "[1300]\ttraining's auc: 0.834926\tvalid_1's auc: 0.786647\n",
      "[1400]\ttraining's auc: 0.837888\tvalid_1's auc: 0.787234\n",
      "[1500]\ttraining's auc: 0.840652\tvalid_1's auc: 0.787763\n",
      "[1600]\ttraining's auc: 0.843374\tvalid_1's auc: 0.788236\n",
      "[1700]\ttraining's auc: 0.845992\tvalid_1's auc: 0.788668\n",
      "[1800]\ttraining's auc: 0.8486\tvalid_1's auc: 0.788957\n",
      "[1900]\ttraining's auc: 0.851222\tvalid_1's auc: 0.789224\n",
      "[2000]\ttraining's auc: 0.853609\tvalid_1's auc: 0.789459\n",
      "[2100]\ttraining's auc: 0.855926\tvalid_1's auc: 0.789668\n",
      "[2200]\ttraining's auc: 0.858131\tvalid_1's auc: 0.789872\n",
      "[2300]\ttraining's auc: 0.86035\tvalid_1's auc: 0.790116\n",
      "[2400]\ttraining's auc: 0.862426\tvalid_1's auc: 0.790244\n",
      "[2500]\ttraining's auc: 0.864591\tvalid_1's auc: 0.790434\n",
      "[2600]\ttraining's auc: 0.866678\tvalid_1's auc: 0.790589\n",
      "[2700]\ttraining's auc: 0.868689\tvalid_1's auc: 0.790726\n",
      "[2800]\ttraining's auc: 0.870514\tvalid_1's auc: 0.79096\n",
      "[2900]\ttraining's auc: 0.872447\tvalid_1's auc: 0.791028\n",
      "[3000]\ttraining's auc: 0.874319\tvalid_1's auc: 0.791172\n",
      "[3100]\ttraining's auc: 0.876138\tvalid_1's auc: 0.791214\n",
      "[3200]\ttraining's auc: 0.877958\tvalid_1's auc: 0.79128\n",
      "[3300]\ttraining's auc: 0.879752\tvalid_1's auc: 0.791283\n",
      "[3400]\ttraining's auc: 0.881483\tvalid_1's auc: 0.791316\n",
      "[3500]\ttraining's auc: 0.88329\tvalid_1's auc: 0.791358\n",
      "[3600]\ttraining's auc: 0.885055\tvalid_1's auc: 0.791334\n",
      "[3700]\ttraining's auc: 0.886753\tvalid_1's auc: 0.791345\n",
      "[3800]\ttraining's auc: 0.888377\tvalid_1's auc: 0.791368\n",
      "[3900]\ttraining's auc: 0.889876\tvalid_1's auc: 0.791382\n",
      "[4000]\ttraining's auc: 0.891426\tvalid_1's auc: 0.791331\n",
      "[4100]\ttraining's auc: 0.893008\tvalid_1's auc: 0.791311\n",
      "[4200]\ttraining's auc: 0.894569\tvalid_1's auc: 0.791269\n",
      "Early stopping, best iteration is:\n",
      "[3915]\ttraining's auc: 0.890151\tvalid_1's auc: 0.791393\n",
      "Fold  5 Train AUC : 0.890149\n",
      "Fold  5 Test AUC : 0.791392\n"
     ]
    }
   ],
   "source": [
    "oof_train, oof_test = get_lgb_oof(data, y, test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_xgb_oof(x_train, y_train, x_test):\n",
    "    oof_train = np.zeros((x_train.shape[0],))\n",
    "    oof_test = np.zeros((x_test.shape[0],))\n",
    "    oof_test_skf = np.empty((NFOLDS, x_test.shape[0]))\n",
    "\n",
    "    for i, (train_index, test_index) in enumerate(folds.split(x_train)):\n",
    "        x_tr = x_train[feats].iloc[train_index]\n",
    "        y_tr = y_train.iloc[train_index]\n",
    "        x_te = x_train[feats].iloc[test_index]\n",
    "        y_te = y_train.iloc[test_index]\n",
    "        clf = XGBClassifier(learning_rate =0.01,\n",
    "                            n_estimators=10000, \n",
    "                            max_depth=4, \n",
    "                            min_child_weight=5, \n",
    "                            subsample=0.8, \n",
    "                            colsample_bytree=0.8,\n",
    "                            objective= 'binary:logistic', \n",
    "                            scale_pos_weight=2.5, \n",
    "                            random_state=SEED,\n",
    "                            reg_lambda = 1.2,\n",
    "                            tree_method='gpu_hist')\n",
    "\n",
    "        clf.fit(x_tr, y_tr, eval_set=[(x_tr, y_tr), (x_te, y_te)], \n",
    "                eval_metric= 'auc', verbose= 100, early_stopping_rounds= 300)\n",
    "\n",
    "        oof_train[test_index] = clf.predict_proba(x_te)[:,1]\n",
    "        oof_test_skf[i, :] = clf.predict_proba(x_test[feats])[:,1]\n",
    "        print('Fold %2d Train AUC : %.6f' % (i + 1, roc_auc_score(y_tr, clf.predict_proba(x_tr)[:,1])))\n",
    "        print('Fold %2d Test AUC : %.6f' % (i + 1, roc_auc_score(y_te, oof_train[test_index])))\n",
    "        del clf,x_tr,y_tr,x_te,y_te\n",
    "        gc.collect()\n",
    "    oof_test[:] = oof_test_skf.mean(axis=0)\n",
    "    return oof_train.reshape(-1, 1), oof_test.reshape(-1, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0]\tvalidation_0-auc:0.703699\tvalidation_1-auc:0.698048\n",
      "Multiple eval metrics have been passed: 'validation_1-auc' will be used for early stopping.\n",
      "\n",
      "Will train until validation_1-auc hasn't improved in 300 rounds.\n",
      "[100]\tvalidation_0-auc:0.737183\tvalidation_1-auc:0.729401\n",
      "[200]\tvalidation_0-auc:0.75055\tvalidation_1-auc:0.740859\n",
      "[300]\tvalidation_0-auc:0.76383\tvalidation_1-auc:0.752257\n",
      "[400]\tvalidation_0-auc:0.774797\tvalidation_1-auc:0.761078\n",
      "[500]\tvalidation_0-auc:0.782732\tvalidation_1-auc:0.767093\n",
      "[600]\tvalidation_0-auc:0.788767\tvalidation_1-auc:0.771541\n",
      "[700]\tvalidation_0-auc:0.793647\tvalidation_1-auc:0.774895\n",
      "[800]\tvalidation_0-auc:0.797645\tvalidation_1-auc:0.777439\n",
      "[900]\tvalidation_0-auc:0.801248\tvalidation_1-auc:0.779569\n",
      "[1000]\tvalidation_0-auc:0.804457\tvalidation_1-auc:0.781212\n",
      "[1100]\tvalidation_0-auc:0.807377\tvalidation_1-auc:0.78267\n",
      "[1200]\tvalidation_0-auc:0.809981\tvalidation_1-auc:0.783821\n",
      "[1300]\tvalidation_0-auc:0.812383\tvalidation_1-auc:0.784832\n",
      "[1400]\tvalidation_0-auc:0.814687\tvalidation_1-auc:0.785633\n",
      "[1500]\tvalidation_0-auc:0.816851\tvalidation_1-auc:0.786357\n",
      "[1600]\tvalidation_0-auc:0.818925\tvalidation_1-auc:0.787014\n",
      "[1700]\tvalidation_0-auc:0.82096\tvalidation_1-auc:0.787652\n",
      "[1800]\tvalidation_0-auc:0.822898\tvalidation_1-auc:0.788266\n",
      "[1900]\tvalidation_0-auc:0.824744\tvalidation_1-auc:0.788811\n",
      "[2000]\tvalidation_0-auc:0.826573\tvalidation_1-auc:0.789308\n",
      "[2100]\tvalidation_0-auc:0.828421\tvalidation_1-auc:0.789754\n",
      "[2200]\tvalidation_0-auc:0.830121\tvalidation_1-auc:0.790102\n",
      "[2300]\tvalidation_0-auc:0.831825\tvalidation_1-auc:0.790408\n",
      "[2400]\tvalidation_0-auc:0.833456\tvalidation_1-auc:0.790787\n",
      "[2500]\tvalidation_0-auc:0.835041\tvalidation_1-auc:0.791113\n",
      "[2600]\tvalidation_0-auc:0.83672\tvalidation_1-auc:0.791432\n",
      "[2700]\tvalidation_0-auc:0.838279\tvalidation_1-auc:0.791678\n",
      "[2800]\tvalidation_0-auc:0.839755\tvalidation_1-auc:0.791955\n",
      "[2900]\tvalidation_0-auc:0.841215\tvalidation_1-auc:0.792233\n",
      "[3000]\tvalidation_0-auc:0.842623\tvalidation_1-auc:0.792371\n",
      "[3100]\tvalidation_0-auc:0.844046\tvalidation_1-auc:0.792599\n",
      "[3200]\tvalidation_0-auc:0.845436\tvalidation_1-auc:0.79278\n",
      "[3300]\tvalidation_0-auc:0.846879\tvalidation_1-auc:0.792983\n",
      "[3400]\tvalidation_0-auc:0.848181\tvalidation_1-auc:0.793063\n",
      "[3500]\tvalidation_0-auc:0.849446\tvalidation_1-auc:0.793235\n",
      "[3600]\tvalidation_0-auc:0.850739\tvalidation_1-auc:0.793393\n",
      "[3700]\tvalidation_0-auc:0.852032\tvalidation_1-auc:0.793515\n",
      "[3800]\tvalidation_0-auc:0.853308\tvalidation_1-auc:0.79365\n",
      "[3900]\tvalidation_0-auc:0.854552\tvalidation_1-auc:0.7938\n",
      "[4000]\tvalidation_0-auc:0.855765\tvalidation_1-auc:0.793873\n",
      "[4100]\tvalidation_0-auc:0.857006\tvalidation_1-auc:0.794006\n",
      "[4200]\tvalidation_0-auc:0.858229\tvalidation_1-auc:0.794104\n",
      "[4300]\tvalidation_0-auc:0.859475\tvalidation_1-auc:0.79415\n",
      "[4400]\tvalidation_0-auc:0.860667\tvalidation_1-auc:0.794253\n",
      "[4500]\tvalidation_0-auc:0.861867\tvalidation_1-auc:0.794373\n",
      "[4600]\tvalidation_0-auc:0.863036\tvalidation_1-auc:0.794443\n",
      "[4700]\tvalidation_0-auc:0.864194\tvalidation_1-auc:0.794439\n",
      "[4800]\tvalidation_0-auc:0.86533\tvalidation_1-auc:0.794408\n",
      "[4900]\tvalidation_0-auc:0.866433\tvalidation_1-auc:0.794463\n",
      "Stopping. Best iteration:\n",
      "[4626]\tvalidation_0-auc:0.863346\tvalidation_1-auc:0.794468\n",
      "\n",
      "Fold  1 Train AUC : 0.866721\n",
      "Fold  1 Test AUC : 0.794463\n",
      "[0]\tvalidation_0-auc:0.702022\tvalidation_1-auc:0.702942\n",
      "Multiple eval metrics have been passed: 'validation_1-auc' will be used for early stopping.\n",
      "\n",
      "Will train until validation_1-auc hasn't improved in 300 rounds.\n",
      "[100]\tvalidation_0-auc:0.736197\tvalidation_1-auc:0.733683\n",
      "[200]\tvalidation_0-auc:0.749485\tvalidation_1-auc:0.744825\n",
      "[300]\tvalidation_0-auc:0.76223\tvalidation_1-auc:0.756274\n",
      "[400]\tvalidation_0-auc:0.773303\tvalidation_1-auc:0.766211\n",
      "[500]\tvalidation_0-auc:0.781445\tvalidation_1-auc:0.772683\n",
      "[600]\tvalidation_0-auc:0.787441\tvalidation_1-auc:0.776937\n",
      "[700]\tvalidation_0-auc:0.792535\tvalidation_1-auc:0.780543\n",
      "[800]\tvalidation_0-auc:0.796591\tvalidation_1-auc:0.782859\n",
      "[900]\tvalidation_0-auc:0.800091\tvalidation_1-auc:0.784817\n",
      "[1000]\tvalidation_0-auc:0.803217\tvalidation_1-auc:0.786445\n",
      "[1100]\tvalidation_0-auc:0.806072\tvalidation_1-auc:0.787768\n",
      "[1200]\tvalidation_0-auc:0.808601\tvalidation_1-auc:0.78881\n",
      "[1300]\tvalidation_0-auc:0.811098\tvalidation_1-auc:0.789869\n",
      "[1400]\tvalidation_0-auc:0.813416\tvalidation_1-auc:0.790721\n",
      "[1500]\tvalidation_0-auc:0.815636\tvalidation_1-auc:0.791446\n",
      "[1600]\tvalidation_0-auc:0.817759\tvalidation_1-auc:0.792106\n",
      "[1700]\tvalidation_0-auc:0.819789\tvalidation_1-auc:0.792568\n",
      "[1800]\tvalidation_0-auc:0.821659\tvalidation_1-auc:0.792969\n",
      "[1900]\tvalidation_0-auc:0.823659\tvalidation_1-auc:0.793503\n",
      "[2000]\tvalidation_0-auc:0.825498\tvalidation_1-auc:0.793851\n",
      "[2100]\tvalidation_0-auc:0.827352\tvalidation_1-auc:0.794283\n",
      "[2200]\tvalidation_0-auc:0.829122\tvalidation_1-auc:0.794655\n",
      "[2300]\tvalidation_0-auc:0.830835\tvalidation_1-auc:0.795066\n",
      "[2400]\tvalidation_0-auc:0.832501\tvalidation_1-auc:0.795321\n",
      "[2500]\tvalidation_0-auc:0.834133\tvalidation_1-auc:0.79557\n",
      "[2600]\tvalidation_0-auc:0.835771\tvalidation_1-auc:0.795883\n",
      "[2700]\tvalidation_0-auc:0.837272\tvalidation_1-auc:0.79613\n",
      "[2800]\tvalidation_0-auc:0.838807\tvalidation_1-auc:0.796353\n",
      "[2900]\tvalidation_0-auc:0.840253\tvalidation_1-auc:0.796519\n",
      "[3000]\tvalidation_0-auc:0.841743\tvalidation_1-auc:0.796647\n",
      "[3100]\tvalidation_0-auc:0.843151\tvalidation_1-auc:0.796827\n",
      "[3200]\tvalidation_0-auc:0.844546\tvalidation_1-auc:0.797007\n",
      "[3300]\tvalidation_0-auc:0.845936\tvalidation_1-auc:0.797142\n",
      "[3400]\tvalidation_0-auc:0.847276\tvalidation_1-auc:0.797323\n",
      "[3500]\tvalidation_0-auc:0.848565\tvalidation_1-auc:0.797447\n",
      "[3600]\tvalidation_0-auc:0.84995\tvalidation_1-auc:0.797552\n",
      "[3700]\tvalidation_0-auc:0.851196\tvalidation_1-auc:0.797708\n",
      "[3800]\tvalidation_0-auc:0.852481\tvalidation_1-auc:0.797816\n",
      "[3900]\tvalidation_0-auc:0.853704\tvalidation_1-auc:0.797934\n",
      "[4000]\tvalidation_0-auc:0.854935\tvalidation_1-auc:0.798044\n",
      "[4100]\tvalidation_0-auc:0.85613\tvalidation_1-auc:0.798036\n",
      "[4200]\tvalidation_0-auc:0.857276\tvalidation_1-auc:0.798104\n",
      "[4300]\tvalidation_0-auc:0.858431\tvalidation_1-auc:0.798189\n",
      "[4400]\tvalidation_0-auc:0.859591\tvalidation_1-auc:0.798257\n",
      "[4500]\tvalidation_0-auc:0.860769\tvalidation_1-auc:0.798351\n",
      "[4600]\tvalidation_0-auc:0.861938\tvalidation_1-auc:0.798417\n",
      "[4700]\tvalidation_0-auc:0.863145\tvalidation_1-auc:0.798493\n",
      "[4800]\tvalidation_0-auc:0.864161\tvalidation_1-auc:0.7985\n",
      "[4900]\tvalidation_0-auc:0.865308\tvalidation_1-auc:0.798522\n",
      "[5000]\tvalidation_0-auc:0.866361\tvalidation_1-auc:0.798562\n",
      "[5100]\tvalidation_0-auc:0.8675\tvalidation_1-auc:0.7986\n",
      "[5200]\tvalidation_0-auc:0.868633\tvalidation_1-auc:0.798645\n",
      "[5300]\tvalidation_0-auc:0.86972\tvalidation_1-auc:0.798691\n",
      "[5400]\tvalidation_0-auc:0.870801\tvalidation_1-auc:0.79873\n",
      "[5500]\tvalidation_0-auc:0.871881\tvalidation_1-auc:0.798807\n",
      "[5600]\tvalidation_0-auc:0.872913\tvalidation_1-auc:0.798766\n",
      "[5700]\tvalidation_0-auc:0.873942\tvalidation_1-auc:0.798792\n",
      "[5800]\tvalidation_0-auc:0.874946\tvalidation_1-auc:0.798731\n",
      "Stopping. Best iteration:\n",
      "[5528]\tvalidation_0-auc:0.872165\tvalidation_1-auc:0.798837\n",
      "\n",
      "Fold  2 Train AUC : 0.875222\n",
      "Fold  2 Test AUC : 0.798729\n",
      "[0]\tvalidation_0-auc:0.702486\tvalidation_1-auc:0.702629\n",
      "Multiple eval metrics have been passed: 'validation_1-auc' will be used for early stopping.\n",
      "\n",
      "Will train until validation_1-auc hasn't improved in 300 rounds.\n",
      "[100]\tvalidation_0-auc:0.736044\tvalidation_1-auc:0.733373\n",
      "[200]\tvalidation_0-auc:0.749711\tvalidation_1-auc:0.74466\n",
      "[300]\tvalidation_0-auc:0.762985\tvalidation_1-auc:0.75515\n",
      "[400]\tvalidation_0-auc:0.774052\tvalidation_1-auc:0.763675\n",
      "[500]\tvalidation_0-auc:0.782175\tvalidation_1-auc:0.769657\n",
      "[600]\tvalidation_0-auc:0.788327\tvalidation_1-auc:0.774012\n",
      "[700]\tvalidation_0-auc:0.793381\tvalidation_1-auc:0.777189\n",
      "[800]\tvalidation_0-auc:0.797681\tvalidation_1-auc:0.779503\n",
      "[900]\tvalidation_0-auc:0.801299\tvalidation_1-auc:0.781247\n",
      "[1000]\tvalidation_0-auc:0.80449\tvalidation_1-auc:0.782653\n",
      "[1100]\tvalidation_0-auc:0.80744\tvalidation_1-auc:0.783984\n",
      "[1200]\tvalidation_0-auc:0.810017\tvalidation_1-auc:0.784985\n",
      "[1300]\tvalidation_0-auc:0.812487\tvalidation_1-auc:0.785826\n",
      "[1400]\tvalidation_0-auc:0.814726\tvalidation_1-auc:0.78661\n",
      "[1500]\tvalidation_0-auc:0.817018\tvalidation_1-auc:0.7873\n",
      "[1600]\tvalidation_0-auc:0.819073\tvalidation_1-auc:0.787956\n",
      "[1700]\tvalidation_0-auc:0.821\tvalidation_1-auc:0.788441\n",
      "[1800]\tvalidation_0-auc:0.822987\tvalidation_1-auc:0.789051\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1900]\tvalidation_0-auc:0.824817\tvalidation_1-auc:0.789444\n",
      "[2000]\tvalidation_0-auc:0.82671\tvalidation_1-auc:0.789795\n",
      "[2100]\tvalidation_0-auc:0.828422\tvalidation_1-auc:0.790154\n",
      "[2200]\tvalidation_0-auc:0.830248\tvalidation_1-auc:0.790486\n",
      "[2300]\tvalidation_0-auc:0.831996\tvalidation_1-auc:0.790851\n",
      "[2400]\tvalidation_0-auc:0.83363\tvalidation_1-auc:0.791045\n",
      "[2500]\tvalidation_0-auc:0.835247\tvalidation_1-auc:0.791279\n",
      "[2600]\tvalidation_0-auc:0.836822\tvalidation_1-auc:0.7914\n",
      "[2700]\tvalidation_0-auc:0.838356\tvalidation_1-auc:0.791559\n",
      "[2800]\tvalidation_0-auc:0.839828\tvalidation_1-auc:0.79174\n",
      "[2900]\tvalidation_0-auc:0.841405\tvalidation_1-auc:0.791888\n",
      "[3000]\tvalidation_0-auc:0.842801\tvalidation_1-auc:0.791974\n",
      "[3100]\tvalidation_0-auc:0.844239\tvalidation_1-auc:0.792019\n",
      "[3200]\tvalidation_0-auc:0.845616\tvalidation_1-auc:0.79215\n",
      "[3300]\tvalidation_0-auc:0.84696\tvalidation_1-auc:0.792279\n",
      "[3400]\tvalidation_0-auc:0.848375\tvalidation_1-auc:0.792363\n",
      "[3500]\tvalidation_0-auc:0.849724\tvalidation_1-auc:0.792474\n",
      "[3600]\tvalidation_0-auc:0.851102\tvalidation_1-auc:0.792583\n",
      "[3700]\tvalidation_0-auc:0.852387\tvalidation_1-auc:0.792688\n",
      "[3800]\tvalidation_0-auc:0.853633\tvalidation_1-auc:0.792794\n",
      "[3900]\tvalidation_0-auc:0.854904\tvalidation_1-auc:0.792858\n",
      "[4000]\tvalidation_0-auc:0.856102\tvalidation_1-auc:0.792949\n",
      "[4100]\tvalidation_0-auc:0.857366\tvalidation_1-auc:0.79298\n",
      "[4200]\tvalidation_0-auc:0.858614\tvalidation_1-auc:0.793032\n",
      "[4300]\tvalidation_0-auc:0.859763\tvalidation_1-auc:0.793087\n",
      "[4400]\tvalidation_0-auc:0.860981\tvalidation_1-auc:0.793167\n",
      "[4500]\tvalidation_0-auc:0.862201\tvalidation_1-auc:0.793213\n",
      "[4600]\tvalidation_0-auc:0.863279\tvalidation_1-auc:0.793251\n",
      "[4700]\tvalidation_0-auc:0.864402\tvalidation_1-auc:0.793288\n",
      "[4800]\tvalidation_0-auc:0.865586\tvalidation_1-auc:0.793356\n",
      "[4900]\tvalidation_0-auc:0.866648\tvalidation_1-auc:0.793373\n",
      "[5000]\tvalidation_0-auc:0.867719\tvalidation_1-auc:0.793318\n",
      "[5100]\tvalidation_0-auc:0.86885\tvalidation_1-auc:0.793344\n",
      "[5200]\tvalidation_0-auc:0.869975\tvalidation_1-auc:0.79342\n",
      "[5300]\tvalidation_0-auc:0.871094\tvalidation_1-auc:0.793485\n",
      "[5400]\tvalidation_0-auc:0.872135\tvalidation_1-auc:0.793525\n",
      "[5500]\tvalidation_0-auc:0.873168\tvalidation_1-auc:0.793561\n",
      "[5600]\tvalidation_0-auc:0.874173\tvalidation_1-auc:0.793608\n",
      "[5700]\tvalidation_0-auc:0.875297\tvalidation_1-auc:0.793621\n",
      "[5800]\tvalidation_0-auc:0.876257\tvalidation_1-auc:0.793601\n",
      "[5900]\tvalidation_0-auc:0.87723\tvalidation_1-auc:0.79363\n",
      "[6000]\tvalidation_0-auc:0.87823\tvalidation_1-auc:0.793614\n",
      "[6100]\tvalidation_0-auc:0.879271\tvalidation_1-auc:0.793611\n",
      "[6200]\tvalidation_0-auc:0.880267\tvalidation_1-auc:0.793592\n",
      "[6300]\tvalidation_0-auc:0.881277\tvalidation_1-auc:0.793584\n",
      "Stopping. Best iteration:\n",
      "[6060]\tvalidation_0-auc:0.878886\tvalidation_1-auc:0.793647\n",
      "\n",
      "Fold  3 Train AUC : 0.881894\n",
      "Fold  3 Test AUC : 0.793534\n",
      "[0]\tvalidation_0-auc:0.704249\tvalidation_1-auc:0.702765\n",
      "Multiple eval metrics have been passed: 'validation_1-auc' will be used for early stopping.\n",
      "\n",
      "Will train until validation_1-auc hasn't improved in 300 rounds.\n",
      "[100]\tvalidation_0-auc:0.736866\tvalidation_1-auc:0.731532\n",
      "[200]\tvalidation_0-auc:0.750089\tvalidation_1-auc:0.74306\n",
      "[300]\tvalidation_0-auc:0.762852\tvalidation_1-auc:0.754428\n",
      "[400]\tvalidation_0-auc:0.773829\tvalidation_1-auc:0.763764\n",
      "[500]\tvalidation_0-auc:0.782137\tvalidation_1-auc:0.770072\n",
      "[600]\tvalidation_0-auc:0.788433\tvalidation_1-auc:0.774363\n",
      "[700]\tvalidation_0-auc:0.793384\tvalidation_1-auc:0.777492\n",
      "[800]\tvalidation_0-auc:0.797464\tvalidation_1-auc:0.779791\n",
      "[900]\tvalidation_0-auc:0.801\tvalidation_1-auc:0.781604\n",
      "[1000]\tvalidation_0-auc:0.804198\tvalidation_1-auc:0.783194\n",
      "[1100]\tvalidation_0-auc:0.80699\tvalidation_1-auc:0.784434\n",
      "[1200]\tvalidation_0-auc:0.809628\tvalidation_1-auc:0.785564\n",
      "[1300]\tvalidation_0-auc:0.812057\tvalidation_1-auc:0.786553\n",
      "[1400]\tvalidation_0-auc:0.814406\tvalidation_1-auc:0.787456\n",
      "[1500]\tvalidation_0-auc:0.816519\tvalidation_1-auc:0.788066\n",
      "[1600]\tvalidation_0-auc:0.818579\tvalidation_1-auc:0.788755\n",
      "[1700]\tvalidation_0-auc:0.82056\tvalidation_1-auc:0.789255\n",
      "[1800]\tvalidation_0-auc:0.822589\tvalidation_1-auc:0.789791\n",
      "[1900]\tvalidation_0-auc:0.824534\tvalidation_1-auc:0.790258\n",
      "[2000]\tvalidation_0-auc:0.82634\tvalidation_1-auc:0.790641\n",
      "[2100]\tvalidation_0-auc:0.828097\tvalidation_1-auc:0.791046\n",
      "[2200]\tvalidation_0-auc:0.829827\tvalidation_1-auc:0.791387\n",
      "[2300]\tvalidation_0-auc:0.831508\tvalidation_1-auc:0.791665\n",
      "[2400]\tvalidation_0-auc:0.833172\tvalidation_1-auc:0.791981\n",
      "[2500]\tvalidation_0-auc:0.834888\tvalidation_1-auc:0.792213\n",
      "[2600]\tvalidation_0-auc:0.836517\tvalidation_1-auc:0.792513\n",
      "[2700]\tvalidation_0-auc:0.838056\tvalidation_1-auc:0.792686\n",
      "[2800]\tvalidation_0-auc:0.839582\tvalidation_1-auc:0.792929\n",
      "[2900]\tvalidation_0-auc:0.841097\tvalidation_1-auc:0.79309\n",
      "[3000]\tvalidation_0-auc:0.842662\tvalidation_1-auc:0.79332\n",
      "[3100]\tvalidation_0-auc:0.844057\tvalidation_1-auc:0.793484\n",
      "[3200]\tvalidation_0-auc:0.84538\tvalidation_1-auc:0.793656\n",
      "[3300]\tvalidation_0-auc:0.846757\tvalidation_1-auc:0.793832\n",
      "[3400]\tvalidation_0-auc:0.848075\tvalidation_1-auc:0.793997\n",
      "[3500]\tvalidation_0-auc:0.849443\tvalidation_1-auc:0.794109\n",
      "[3600]\tvalidation_0-auc:0.850813\tvalidation_1-auc:0.794223\n",
      "[3700]\tvalidation_0-auc:0.852056\tvalidation_1-auc:0.794292\n",
      "[3800]\tvalidation_0-auc:0.853271\tvalidation_1-auc:0.794358\n",
      "[3900]\tvalidation_0-auc:0.854548\tvalidation_1-auc:0.794462\n",
      "[4000]\tvalidation_0-auc:0.855863\tvalidation_1-auc:0.794549\n",
      "[4100]\tvalidation_0-auc:0.857131\tvalidation_1-auc:0.794675\n",
      "[4200]\tvalidation_0-auc:0.858311\tvalidation_1-auc:0.794776\n",
      "[4300]\tvalidation_0-auc:0.859463\tvalidation_1-auc:0.794837\n",
      "[4400]\tvalidation_0-auc:0.860705\tvalidation_1-auc:0.794898\n",
      "[4500]\tvalidation_0-auc:0.861872\tvalidation_1-auc:0.794988\n",
      "[4600]\tvalidation_0-auc:0.863076\tvalidation_1-auc:0.79507\n",
      "[4700]\tvalidation_0-auc:0.864272\tvalidation_1-auc:0.795168\n",
      "[4800]\tvalidation_0-auc:0.865468\tvalidation_1-auc:0.795258\n",
      "[4900]\tvalidation_0-auc:0.866577\tvalidation_1-auc:0.795301\n",
      "[5000]\tvalidation_0-auc:0.867757\tvalidation_1-auc:0.795335\n",
      "[5100]\tvalidation_0-auc:0.868924\tvalidation_1-auc:0.795373\n",
      "[5200]\tvalidation_0-auc:0.870053\tvalidation_1-auc:0.795382\n",
      "[5300]\tvalidation_0-auc:0.871115\tvalidation_1-auc:0.795391\n",
      "[5400]\tvalidation_0-auc:0.872226\tvalidation_1-auc:0.795393\n",
      "[5500]\tvalidation_0-auc:0.873307\tvalidation_1-auc:0.795449\n",
      "[5600]\tvalidation_0-auc:0.874321\tvalidation_1-auc:0.795517\n",
      "[5700]\tvalidation_0-auc:0.875348\tvalidation_1-auc:0.795557\n",
      "[5800]\tvalidation_0-auc:0.876366\tvalidation_1-auc:0.795568\n",
      "[5900]\tvalidation_0-auc:0.877445\tvalidation_1-auc:0.795538\n",
      "[6000]\tvalidation_0-auc:0.878452\tvalidation_1-auc:0.795525\n",
      "Stopping. Best iteration:\n",
      "[5725]\tvalidation_0-auc:0.87562\tvalidation_1-auc:0.79559\n",
      "\n",
      "Fold  4 Train AUC : 0.878688\n",
      "Fold  4 Test AUC : 0.795521\n",
      "[0]\tvalidation_0-auc:0.702458\tvalidation_1-auc:0.697537\n",
      "Multiple eval metrics have been passed: 'validation_1-auc' will be used for early stopping.\n",
      "\n",
      "Will train until validation_1-auc hasn't improved in 300 rounds.\n",
      "[100]\tvalidation_0-auc:0.738547\tvalidation_1-auc:0.728274\n",
      "[200]\tvalidation_0-auc:0.751692\tvalidation_1-auc:0.738989\n",
      "[300]\tvalidation_0-auc:0.764513\tvalidation_1-auc:0.74907\n",
      "[400]\tvalidation_0-auc:0.775672\tvalidation_1-auc:0.757703\n",
      "[500]\tvalidation_0-auc:0.783691\tvalidation_1-auc:0.763447\n",
      "[600]\tvalidation_0-auc:0.789917\tvalidation_1-auc:0.767736\n",
      "[700]\tvalidation_0-auc:0.7947\tvalidation_1-auc:0.770898\n",
      "[800]\tvalidation_0-auc:0.798867\tvalidation_1-auc:0.773481\n",
      "[900]\tvalidation_0-auc:0.802241\tvalidation_1-auc:0.77558\n",
      "[1000]\tvalidation_0-auc:0.805396\tvalidation_1-auc:0.777308\n",
      "[1100]\tvalidation_0-auc:0.808309\tvalidation_1-auc:0.778844\n",
      "[1200]\tvalidation_0-auc:0.810844\tvalidation_1-auc:0.780054\n",
      "[1300]\tvalidation_0-auc:0.813344\tvalidation_1-auc:0.781095\n",
      "[1400]\tvalidation_0-auc:0.815665\tvalidation_1-auc:0.78199\n",
      "[1500]\tvalidation_0-auc:0.817789\tvalidation_1-auc:0.782839\n",
      "[1600]\tvalidation_0-auc:0.819852\tvalidation_1-auc:0.783505\n",
      "[1700]\tvalidation_0-auc:0.821854\tvalidation_1-auc:0.784073\n",
      "[1800]\tvalidation_0-auc:0.823834\tvalidation_1-auc:0.7847\n",
      "[1900]\tvalidation_0-auc:0.82578\tvalidation_1-auc:0.785207\n",
      "[2000]\tvalidation_0-auc:0.827577\tvalidation_1-auc:0.7857\n",
      "[2100]\tvalidation_0-auc:0.829329\tvalidation_1-auc:0.786098\n",
      "[2200]\tvalidation_0-auc:0.831021\tvalidation_1-auc:0.786373\n",
      "[2300]\tvalidation_0-auc:0.832659\tvalidation_1-auc:0.786648\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2400]\tvalidation_0-auc:0.834278\tvalidation_1-auc:0.786948\n",
      "[2500]\tvalidation_0-auc:0.835916\tvalidation_1-auc:0.787174\n",
      "[2600]\tvalidation_0-auc:0.837499\tvalidation_1-auc:0.787418\n",
      "[2700]\tvalidation_0-auc:0.839009\tvalidation_1-auc:0.787704\n",
      "[2800]\tvalidation_0-auc:0.840396\tvalidation_1-auc:0.787872\n",
      "[2900]\tvalidation_0-auc:0.841838\tvalidation_1-auc:0.788115\n",
      "[3000]\tvalidation_0-auc:0.843394\tvalidation_1-auc:0.788339\n",
      "[3100]\tvalidation_0-auc:0.844799\tvalidation_1-auc:0.788479\n",
      "[3200]\tvalidation_0-auc:0.846216\tvalidation_1-auc:0.788605\n",
      "[3300]\tvalidation_0-auc:0.847572\tvalidation_1-auc:0.788691\n",
      "[3400]\tvalidation_0-auc:0.848931\tvalidation_1-auc:0.788839\n",
      "[3500]\tvalidation_0-auc:0.850272\tvalidation_1-auc:0.788952\n",
      "[3600]\tvalidation_0-auc:0.851618\tvalidation_1-auc:0.789035\n",
      "[3700]\tvalidation_0-auc:0.852913\tvalidation_1-auc:0.789187\n",
      "[3800]\tvalidation_0-auc:0.854191\tvalidation_1-auc:0.789334\n",
      "[3900]\tvalidation_0-auc:0.855435\tvalidation_1-auc:0.78942\n",
      "[4000]\tvalidation_0-auc:0.8567\tvalidation_1-auc:0.789487\n",
      "[4100]\tvalidation_0-auc:0.857912\tvalidation_1-auc:0.789558\n",
      "[4200]\tvalidation_0-auc:0.859083\tvalidation_1-auc:0.78957\n",
      "[4300]\tvalidation_0-auc:0.860245\tvalidation_1-auc:0.78963\n",
      "[4400]\tvalidation_0-auc:0.861443\tvalidation_1-auc:0.789684\n",
      "[4500]\tvalidation_0-auc:0.862593\tvalidation_1-auc:0.789763\n",
      "[4600]\tvalidation_0-auc:0.863796\tvalidation_1-auc:0.789846\n",
      "[4700]\tvalidation_0-auc:0.864938\tvalidation_1-auc:0.789933\n",
      "[4800]\tvalidation_0-auc:0.866021\tvalidation_1-auc:0.789992\n",
      "[4900]\tvalidation_0-auc:0.867188\tvalidation_1-auc:0.790077\n",
      "[5000]\tvalidation_0-auc:0.868258\tvalidation_1-auc:0.790068\n",
      "[5100]\tvalidation_0-auc:0.869368\tvalidation_1-auc:0.790115\n",
      "[5200]\tvalidation_0-auc:0.870467\tvalidation_1-auc:0.790175\n",
      "[5300]\tvalidation_0-auc:0.871514\tvalidation_1-auc:0.79025\n",
      "[5400]\tvalidation_0-auc:0.872558\tvalidation_1-auc:0.790263\n",
      "[5500]\tvalidation_0-auc:0.873635\tvalidation_1-auc:0.790319\n",
      "[5600]\tvalidation_0-auc:0.874737\tvalidation_1-auc:0.790339\n",
      "[5700]\tvalidation_0-auc:0.875807\tvalidation_1-auc:0.790401\n",
      "[5800]\tvalidation_0-auc:0.876802\tvalidation_1-auc:0.790388\n",
      "[5900]\tvalidation_0-auc:0.877786\tvalidation_1-auc:0.790375\n",
      "Stopping. Best iteration:\n",
      "[5691]\tvalidation_0-auc:0.875712\tvalidation_1-auc:0.790408\n",
      "\n",
      "Fold  5 Train AUC : 0.878746\n",
      "Fold  5 Test AUC : 0.790390\n"
     ]
    }
   ],
   "source": [
    "oof_train, oof_test = get_xgb_oof(data, y, test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "oof_train_csv = pd.DataFrame()\n",
    "oof_train_csv['ID'] = data.SK_ID_CURR\n",
    "oof_train_csv['xgb_oof_train'] = oof_train\n",
    "\n",
    "oof_train_csv.to_csv('./output/xgb_oof_train3.csv', index=False)\n",
    "\n",
    "oof_test_csv = pd.DataFrame()\n",
    "oof_test_csv['ID'] = test.SK_ID_CURR\n",
    "oof_test_csv['xgb_oof_test'] = oof_test\n",
    "\n",
    "oof_test_csv.to_csv('./output/xgb_oof_test3.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training until validation scores don't improve for 300 rounds.\n",
      "[100]\ttraining's auc: 0.743163\tvalid_1's auc: 0.732666\n",
      "[200]\ttraining's auc: 0.753149\tvalid_1's auc: 0.740641\n",
      "[300]\ttraining's auc: 0.768547\tvalid_1's auc: 0.752516\n",
      "[400]\ttraining's auc: 0.784311\tvalid_1's auc: 0.764422\n",
      "[500]\ttraining's auc: 0.795702\tvalid_1's auc: 0.77257\n"
     ]
    }
   ],
   "source": [
    "for n_fold, (trn_idx, val_idx) in enumerate(folds.split(data)):\n",
    "    trn_x, trn_y = data[feats].iloc[trn_idx], y.iloc[trn_idx]\n",
    "    val_x, val_y = data[feats].iloc[val_idx], y.iloc[val_idx]\n",
    "    \n",
    "    clf = LGBMClassifier(\n",
    "        n_estimators=10000,\n",
    "        learning_rate=0.01,\n",
    "        num_leaves=31,\n",
    "        colsample_bytree=0.9338690,\n",
    "        subsample=0.97291513,\n",
    "        max_depth=8,\n",
    "        reg_alpha=0.06663684,\n",
    "        reg_lambda=0.0988343,\n",
    "        min_split_gain=0.07916152,\n",
    "        min_child_weight=82.8850799,\n",
    "        random_state=1700,\n",
    "    )\n",
    "    clf.fit(trn_x, trn_y, \n",
    "            eval_set= [(trn_x, trn_y), (val_x, val_y)], \n",
    "            eval_metric='auc', verbose=100, early_stopping_rounds=300  #30\n",
    "           )\n",
    "    \n",
    "    oof_preds[val_idx] = clf.predict_proba(val_x, num_iteration=clf.best_iteration_)[:, 1]\n",
    "    sub_preds += clf.predict_proba(test[feats], num_iteration=clf.best_iteration_)[:, 1] / folds.n_splits\n",
    "    fold_importance_df = pd.DataFrame()\n",
    "    fold_importance_df[\"feature\"] = feats\n",
    "    fold_importance_df[\"importance\"] = clf.feature_importances_\n",
    "    fold_importance_df[\"fold\"] = n_fold + 1\n",
    "    feature_importance_df = pd.concat([feature_importance_df, fold_importance_df], axis=0)\n",
    "    print('Fold %2d AUC : %.6f' % (n_fold + 1, roc_auc_score(val_y, oof_preds[val_idx])))\n",
    "    del clf, trn_x, trn_y, val_x, val_y\n",
    "    gc.collect()\n",
    "    \n",
    "print('Full AUC score %.6f' % roc_auc_score(y, oof_preds)) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training until validation scores don't improve for 300 rounds.\n",
      "[100]\ttraining's auc: 0.742838\tvalid_1's auc: 0.732527\n",
      "[200]\ttraining's auc: 0.753141\tvalid_1's auc: 0.74064\n",
      "[300]\ttraining's auc: 0.768644\tvalid_1's auc: 0.752598\n",
      "[400]\ttraining's auc: 0.7842\tvalid_1's auc: 0.764355\n",
      "[500]\ttraining's auc: 0.795634\tvalid_1's auc: 0.772435\n",
      "[600]\ttraining's auc: 0.803951\tvalid_1's auc: 0.778038\n",
      "[700]\ttraining's auc: 0.810601\tvalid_1's auc: 0.781943\n",
      "[800]\ttraining's auc: 0.816244\tvalid_1's auc: 0.784882\n",
      "[900]\ttraining's auc: 0.821074\tvalid_1's auc: 0.786905\n",
      "[1000]\ttraining's auc: 0.825134\tvalid_1's auc: 0.788235\n",
      "[1100]\ttraining's auc: 0.829091\tvalid_1's auc: 0.78947\n",
      "[1200]\ttraining's auc: 0.832595\tvalid_1's auc: 0.790383\n",
      "[1300]\ttraining's auc: 0.83586\tvalid_1's auc: 0.791198\n",
      "[1400]\ttraining's auc: 0.838794\tvalid_1's auc: 0.79185\n",
      "[1500]\ttraining's auc: 0.841554\tvalid_1's auc: 0.792315\n",
      "[1600]\ttraining's auc: 0.844341\tvalid_1's auc: 0.792798\n",
      "[1700]\ttraining's auc: 0.84701\tvalid_1's auc: 0.793243\n",
      "[1800]\ttraining's auc: 0.849582\tvalid_1's auc: 0.793574\n",
      "[1900]\ttraining's auc: 0.852153\tvalid_1's auc: 0.793905\n",
      "[2000]\ttraining's auc: 0.854512\tvalid_1's auc: 0.79418\n",
      "[2100]\ttraining's auc: 0.856858\tvalid_1's auc: 0.794433\n",
      "[2200]\ttraining's auc: 0.859116\tvalid_1's auc: 0.794628\n",
      "[2300]\ttraining's auc: 0.861281\tvalid_1's auc: 0.794839\n",
      "[2400]\ttraining's auc: 0.863344\tvalid_1's auc: 0.795049\n",
      "[2500]\ttraining's auc: 0.865418\tvalid_1's auc: 0.795171\n",
      "[2600]\ttraining's auc: 0.867448\tvalid_1's auc: 0.795242\n",
      "[2700]\ttraining's auc: 0.86933\tvalid_1's auc: 0.795316\n",
      "[2800]\ttraining's auc: 0.87104\tvalid_1's auc: 0.795471\n",
      "[2900]\ttraining's auc: 0.872975\tvalid_1's auc: 0.795554\n",
      "[3000]\ttraining's auc: 0.874794\tvalid_1's auc: 0.795685\n",
      "[3100]\ttraining's auc: 0.876598\tvalid_1's auc: 0.795711\n",
      "[3200]\ttraining's auc: 0.87846\tvalid_1's auc: 0.795754\n",
      "[3300]\ttraining's auc: 0.880201\tvalid_1's auc: 0.795837\n",
      "[3400]\ttraining's auc: 0.881993\tvalid_1's auc: 0.795828\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-6-1d8c737f0e94>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     18\u001b[0m     clf.fit(trn_x, trn_y, \n\u001b[1;32m     19\u001b[0m             \u001b[0meval_set\u001b[0m\u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtrn_x\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtrn_y\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mval_x\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mval_y\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 20\u001b[0;31m             \u001b[0meval_metric\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m'auc'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mverbose\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m100\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mearly_stopping_rounds\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m300\u001b[0m  \u001b[0;31m#30\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     21\u001b[0m            )\n\u001b[1;32m     22\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/data-mine/lib/python3.5/site-packages/lightgbm/sklearn.py\u001b[0m in \u001b[0;36mfit\u001b[0;34m(self, X, y, sample_weight, init_score, eval_set, eval_names, eval_sample_weight, eval_class_weight, eval_init_score, eval_metric, early_stopping_rounds, verbose, feature_name, categorical_feature, callbacks)\u001b[0m\n\u001b[1;32m    673\u001b[0m                                         \u001b[0mverbose\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mverbose\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeature_name\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mfeature_name\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    674\u001b[0m                                         \u001b[0mcategorical_feature\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mcategorical_feature\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 675\u001b[0;31m                                         callbacks=callbacks)\n\u001b[0m\u001b[1;32m    676\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    677\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/data-mine/lib/python3.5/site-packages/lightgbm/sklearn.py\u001b[0m in \u001b[0;36mfit\u001b[0;34m(self, X, y, sample_weight, init_score, group, eval_set, eval_names, eval_sample_weight, eval_class_weight, eval_init_score, eval_group, eval_metric, early_stopping_rounds, verbose, feature_name, categorical_feature, callbacks)\u001b[0m\n\u001b[1;32m    467\u001b[0m                               \u001b[0mverbose_eval\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mverbose\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeature_name\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mfeature_name\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    468\u001b[0m                               \u001b[0mcategorical_feature\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mcategorical_feature\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 469\u001b[0;31m                               callbacks=callbacks)\n\u001b[0m\u001b[1;32m    470\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    471\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mevals_result\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/data-mine/lib/python3.5/site-packages/lightgbm/engine.py\u001b[0m in \u001b[0;36mtrain\u001b[0;34m(params, train_set, num_boost_round, valid_sets, valid_names, fobj, feval, init_model, feature_name, categorical_feature, early_stopping_rounds, evals_result, verbose_eval, learning_rates, keep_training_booster, callbacks)\u001b[0m\n\u001b[1;32m    199\u001b[0m                                     evaluation_result_list=None))\n\u001b[1;32m    200\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 201\u001b[0;31m         \u001b[0mbooster\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mupdate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfobj\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mfobj\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    202\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    203\u001b[0m         \u001b[0mevaluation_result_list\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/data-mine/lib/python3.5/site-packages/lightgbm/basic.py\u001b[0m in \u001b[0;36mupdate\u001b[0;34m(self, train_set, fobj)\u001b[0m\n\u001b[1;32m   1522\u001b[0m             _safe_call(_LIB.LGBM_BoosterUpdateOneIter(\n\u001b[1;32m   1523\u001b[0m                 \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mhandle\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1524\u001b[0;31m                 ctypes.byref(is_finished)))\n\u001b[0m\u001b[1;32m   1525\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__is_predicted_cur_iter\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;32mFalse\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0m_\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mrange_\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__num_dataset\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1526\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0mis_finished\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalue\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "for n_fold, (trn_idx, val_idx) in enumerate(folds.split(data)):\n",
    "    trn_x, trn_y = data[feats].iloc[trn_idx], y.iloc[trn_idx]\n",
    "    val_x, val_y = data[feats].iloc[val_idx], y.iloc[val_idx]\n",
    "    \n",
    "    clf = LGBMClassifier(\n",
    "        n_estimators=10000,\n",
    "        learning_rate=0.01,\n",
    "        num_leaves=31,\n",
    "        colsample_bytree=0.9338690,\n",
    "        subsample=0.97291513,\n",
    "        max_depth=8,\n",
    "        reg_alpha=0.06663684,\n",
    "        reg_lambda=0.0988343,\n",
    "        min_split_gain=0.07916152,\n",
    "        min_child_weight=82.8850799,\n",
    "        random_state=1700,\n",
    "    )\n",
    "    clf.fit(trn_x, trn_y, \n",
    "            eval_set= [(trn_x, trn_y), (val_x, val_y)], \n",
    "            eval_metric='auc', verbose=100, early_stopping_rounds=300  #30\n",
    "           )\n",
    "    \n",
    "    oof_preds[val_idx] = clf.predict_proba(val_x, num_iteration=clf.best_iteration_)[:, 1]\n",
    "    sub_preds += clf.predict_proba(test[feats], num_iteration=clf.best_iteration_)[:, 1] / folds.n_splits\n",
    "    fold_importance_df = pd.DataFrame()\n",
    "    fold_importance_df[\"feature\"] = feats\n",
    "    fold_importance_df[\"importance\"] = clf.feature_importances_\n",
    "    fold_importance_df[\"fold\"] = n_fold + 1\n",
    "    feature_importance_df = pd.concat([feature_importance_df, fold_importance_df], axis=0)\n",
    "    print('Fold %2d AUC : %.6f' % (n_fold + 1, roc_auc_score(val_y, oof_preds[val_idx])))\n",
    "    del clf, trn_x, trn_y, val_x, val_y\n",
    "    gc.collect()\n",
    "    \n",
    "print('Full AUC score %.6f' % roc_auc_score(y, oof_preds)) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x720 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "cols = feature_importance_df[[\"feature\", \"importance\"]].groupby(\"feature\").mean().sort_values(by=\"importance\", ascending=False)[0:50].index\n",
    "best_features = feature_importance_df.loc[feature_importance_df.feature.isin(cols)]\n",
    "plt.figure(figsize=(12, 10))\n",
    "sns.barplot(x=\"importance\", y=\"feature\", data=best_features.sort_values(by=\"importance\", ascending=False))\n",
    "plt.title('LightGBM Features (avg over folds)')\n",
    "plt.tight_layout()\n",
    "plt.savefig('lgbm_importances-7952.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "test['TARGET'] = sub_preds\n",
    "\n",
    "test[['SK_ID_CURR', 'TARGET']].to_csv('submission7952.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "ename": "KeyError",
     "evalue": "\"['feature' 'importance'] not in index\"",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-7-026dadc1b06f>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mcols\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfeature_importance_df\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m\"feature\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\"importance\"\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mgroupby\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"feature\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmean\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msort_values\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mby\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m\"importance\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mascending\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;36m50\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mindex\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      2\u001b[0m \u001b[0mbest_features\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfeature_importance_df\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mloc\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mfeature_importance_df\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfeature\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0misin\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcols\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0mplt\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfigure\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfigsize\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m12\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m10\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0msns\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbarplot\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m\"importance\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m\"feature\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdata\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mbest_features\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msort_values\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mby\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m\"importance\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mascending\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0mplt\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtitle\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'LightGBM Features (avg over folds)'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/data-mine/lib/python3.5/site-packages/pandas/core/frame.py\u001b[0m in \u001b[0;36m__getitem__\u001b[0;34m(self, key)\u001b[0m\n\u001b[1;32m   2680\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mSeries\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mndarray\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mIndex\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlist\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2681\u001b[0m             \u001b[0;31m# either boolean or fancy integer index\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2682\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_getitem_array\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   2683\u001b[0m         \u001b[0;32melif\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mDataFrame\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2684\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_getitem_frame\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/data-mine/lib/python3.5/site-packages/pandas/core/frame.py\u001b[0m in \u001b[0;36m_getitem_array\u001b[0;34m(self, key)\u001b[0m\n\u001b[1;32m   2724\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_take\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mindexer\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0maxis\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2725\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2726\u001b[0;31m             \u001b[0mindexer\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mloc\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_convert_to_indexer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0maxis\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   2727\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_take\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mindexer\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0maxis\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2728\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/data-mine/lib/python3.5/site-packages/pandas/core/indexing.py\u001b[0m in \u001b[0;36m_convert_to_indexer\u001b[0;34m(self, obj, axis, is_setter)\u001b[0m\n\u001b[1;32m   1325\u001b[0m                 \u001b[0;32mif\u001b[0m \u001b[0mmask\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0many\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1326\u001b[0m                     raise KeyError('{mask} not in index'\n\u001b[0;32m-> 1327\u001b[0;31m                                    .format(mask=objarr[mask]))\n\u001b[0m\u001b[1;32m   1328\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1329\u001b[0m                 \u001b[0;32mreturn\u001b[0m \u001b[0mcom\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_values_from_object\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mindexer\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mKeyError\u001b[0m: \"['feature' 'importance'] not in index\""
     ]
    }
   ],
   "source": [
    "cols = feature_importance_df[[\"feature\", \"importance\"]].groupby(\"feature\").mean().sort_values(by=\"importance\", ascending=False)[0:50].index\n",
    "best_features = feature_importance_df.loc[feature_importance_df.feature.isin(cols)]\n",
    "plt.figure(figsize=(12, 10))\n",
    "sns.barplot(x=\"importance\", y=\"feature\", data=best_features.sort_values(by=\"importance\", ascending=False))\n",
    "plt.title('LightGBM Features (avg over folds)')\n",
    "plt.tight_layout()\n",
    "#plt.savefig('lgbm_importances-7952.png')"
   ]
  },
  {
   "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}