Main.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 93,
   "id": "28204bab",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Suppress warnings \n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "\n",
    "import os, sys, gc, warnings, random\n",
    "\n",
    "import datetime\n",
    "import dateutil.relativedelta\n",
    "\n",
    "# Data manipulation\n",
    "import pandas as pd \n",
    "import numpy as np\n",
    "\n",
    "# Visualization\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "# Machine learning\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.preprocessing import MinMaxScaler, LabelEncoder\n",
    "from sklearn.impute import SimpleImputer\n",
    "from sklearn.metrics import precision_score, recall_score, f1_score, roc_auc_score, precision_recall_curve, roc_curve\n",
    "from sklearn.model_selection import train_test_split, cross_val_score, KFold, StratifiedKFold, GroupKFold\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.preprocessing import quantile_transform\n",
    "import scipy.stats as stats\n",
    "\n",
    "\n",
    "import lightgbm as lgb\n",
    "import optuna\n",
    "\n",
    "from catboost import CatBoostClassifier, Pool\n",
    "from tqdm.notebook import trange, tqdm\n",
    "\n",
    "from IPython.display import display\n",
    "\n",
    "%matplotlib inline\n",
    "\n",
    "pd.options.display.max_rows = 10000\n",
    "pd.options.display.max_columns = 1000\n",
    "pd.options.display.max_colwidth = 1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "id": "8350a66a",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv('./input' + '/train.csv', parse_dates=['order_date'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "id": "10747b24",
   "metadata": {},
   "outputs": [],
   "source": [
    "def seed_everything(seed=0):\n",
    "    random.seed(seed)\n",
    "    os.environ['PYTHONHASHSEED'] = str(seed)\n",
    "    np.random.seed(seed)\n",
    "    \n",
    "SEED = 42\n",
    "seed_everything(SEED)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "id": "76dd3f19",
   "metadata": {},
   "outputs": [],
   "source": [
    "TOTAL_THRES = 300\n",
    "\n",
    "'''\n",
    "    입력인자로 받는 year_month에 대해 고객 ID별로 총 구매액이\n",
    "    구매액 임계값을 넘는지 여부의 binary label을 생성하는 함수\n",
    "'''\n",
    "def generate_label(df, year_month, total_thres=TOTAL_THRES, print_log=False):\n",
    "    \n",
    "    \n",
    "    df = df.copy()\n",
    "    \n",
    "    # year_month에 해당하는 label 데이터 생성\n",
    "    df['year_month'] = df['order_date'].dt.strftime('%Y-%m')\n",
    "    df.reset_index(drop=True, inplace=True)\n",
    "\n",
    "    # year_month 이전 월의 고객 ID 추출\n",
    "    cust = df[df['year_month']<year_month]['customer_id'].unique()\n",
    "    # year_month에 해당하는 데이터 선택\n",
    "    df = df[df['year_month']==year_month]\n",
    "    \n",
    "    # label 데이터프레임 생성\n",
    "    label = pd.DataFrame({'customer_id':cust})\n",
    "    label['year_month'] = year_month\n",
    "    \n",
    "    # year_month에 해당하는 고객 ID의 구매액의 합 계산\n",
    "    grped = df.groupby(['customer_id','year_month'], as_index=False)[['total']].sum()\n",
    "    \n",
    "    # label 데이터프레임과 merge하고 구매액 임계값을 넘었는지 여부로 label 생성\n",
    "    label = label.merge(grped, on=['customer_id','year_month'], how='left')\n",
    "    label['total'].fillna(0.0, inplace=True)\n",
    "    label['label'] = (label['total'] > total_thres).astype(int)\n",
    "\n",
    "    # 고객 ID로 정렬\n",
    "    label = label.sort_values('customer_id').reset_index(drop=True)\n",
    "    if print_log: print(f'{year_month} - final label shape: {label.shape}')\n",
    "    \n",
    "    return label"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "id": "9ce460b9",
   "metadata": {},
   "outputs": [],
   "source": [
    "def print_score(label, pred, prob_thres=0.5):\n",
    "    print('Precision: {:.5f}'.format(precision_score(label, pred>prob_thres)))\n",
    "    print('Recall: {:.5f}'.format(recall_score(label, pred>prob_thres)))\n",
    "    print('F1 Score: {:.5f}'.format(f1_score(label, pred>prob_thres)))\n",
    "    print('ROC AUC Score: {:.5f}'.format(roc_auc_score(label, pred)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "id": "e90c803b",
   "metadata": {},
   "outputs": [],
   "source": [
    "def qt(df):\n",
    "    df = df.copy()\n",
    "    \n",
    "    for col in df.columns:\n",
    "        if df[col].dtype.name == 'float64':\n",
    "            temp = quantile_transform(np.array(df[col]).astype(\"float\").reshape(-1,1), n_quantiles=100, random_state=0, copy=True, output_distribution = \"normal\").reshape(-1)   \n",
    "            df[col] = temp\n",
    "    return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "id": "120f9e91",
   "metadata": {},
   "outputs": [],
   "source": [
    "def feature_preprocessing1(train, test, features, do_imputing=True):\n",
    "    x_tr = train.copy()\n",
    "    x_te = test.copy()\n",
    "    \n",
    "    # 범주형 피처 이름을 저장할 변수\n",
    "    cate_cols = []\n",
    "\n",
    "    # 레이블 인코딩\n",
    "    for f in features:\n",
    "        if x_tr[f].dtype.name == 'object': # 데이터 타입이 object(str)이면 레이블 인코딩\n",
    "            cate_cols.append(f)\n",
    "            le = LabelEncoder()\n",
    "            # train + test 데이터를 합쳐서 레이블 인코딩 함수에 fit\n",
    "            le.fit(list(x_tr[f].values) + list(x_te[f].values))\n",
    "            \n",
    "            # train 데이터 레이블 인코딩 변환 수행\n",
    "            x_tr[f] = le.transform(list(x_tr[f].values))\n",
    "            \n",
    "            # test 데이터 레이블 인코딩 변환 수행\n",
    "            x_te[f] = le.transform(list(x_te[f].values))\n",
    "\n",
    "    print('categorical feature:', cate_cols)\n",
    "\n",
    "    if do_imputing:\n",
    "        # 중위값으로 결측치 채우기\n",
    "        imputer = SimpleImputer(strategy='median')\n",
    "\n",
    "        x_tr[features] = imputer.fit_transform(x_tr[features])\n",
    "        x_te[features] = imputer.transform(x_te[features])\n",
    "    \n",
    "    return x_tr, x_te"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "id": "c387c1d6",
   "metadata": {},
   "outputs": [],
   "source": [
    "def feature_engineering1(df, year_month):\n",
    "    df = df.copy()\n",
    "    \n",
    "    #Time Series diff Feature Generation\n",
    "    df['order_ts'] = df['order_date'].astype(np.int64) // 1e9\n",
    "    df['order_ts_diff'] = df.groupby(['customer_id'])['order_ts'].diff()\n",
    "    df['quantity_diff'] = df.groupby(['customer_id'])['quantity'].diff()\n",
    "    df['price_diff'] = df.groupby(['customer_id'])['price'].diff()\n",
    "    df['total_diff'] = df.groupby(['customer_id'])['total'].diff()\n",
    "    \n",
    "    \n",
    "    #cumsum agg\n",
    "    df['cumsum_total_by_cust_id'] = df.groupby(['customer_id'])['total'].cumsum()\n",
    "    df['cumsum_quantity_by_cust_id'] = df.groupby(['customer_id'])['quantity'].cumsum()\n",
    "    df['cumsum_price_by_cust_id'] = df.groupby(['customer_id'])['price'].cumsum()\n",
    "    \n",
    "    df['cumsum_total_by_prod_id'] = df.groupby(['product_id'])['total'].cumsum()\n",
    "    df['cumsum_quantity_by_prod_id'] = df.groupby(['product_id'])['quantity'].cumsum()\n",
    "    df['cumsum_price_by_prod_id'] = df.groupby(['product_id'])['price'].cumsum()\n",
    "    \n",
    "    df['cumsum_total_by_order_id'] = df.groupby(['order_id'])['total'].cumsum()\n",
    "    df['cumsum_quantity_by_order_id'] = df.groupby(['order_id'])['quantity'].cumsum()\n",
    "    df['cumsum_price_by_order_id'] = df.groupby(['order_id'])['price'].cumsum()    \n",
    "    \n",
    "    # year_month 이전 월 계산\n",
    "    d = datetime.datetime.strptime(year_month, \"%Y-%m\")\n",
    "    prev_ym = d - dateutil.relativedelta.relativedelta(months=1)\n",
    "    prev_ym = prev_ym.strftime('%Y-%m')\n",
    "    \n",
    "    # train, test 데이터 선택\n",
    "    train = df[df['order_date'] < prev_ym]\n",
    "    test = df[df['order_date'] < year_month]    \n",
    "    \n",
    "    # train, test 레이블 데이터 생성  \n",
    "    train_label = generate_label(df, prev_ym)[['customer_id','year_month','label']]\n",
    "    test_label = generate_label(df, year_month)[['customer_id','year_month','label']]\n",
    "\n",
    "    \n",
    "    # group by aggregation 함수 선언\n",
    "    agg_func = ['mean','max','min','sum','count','std','skew']\n",
    "    agg_dict = {\n",
    "        'quantity': agg_func,\n",
    "        'price': agg_func,\n",
    "        'total': agg_func,\n",
    "        'cumsum_total_by_cust_id': agg_func,\n",
    "        'cumsum_quantity_by_cust_id': agg_func,\n",
    "        'cumsum_price_by_cust_id': agg_func,\n",
    "        'cumsum_total_by_prod_id': agg_func,\n",
    "        'cumsum_quantity_by_prod_id': agg_func,\n",
    "        'cumsum_price_by_prod_id': agg_func,\n",
    "        'cumsum_total_by_order_id': agg_func,\n",
    "        'cumsum_quantity_by_order_id': agg_func,\n",
    "        'cumsum_price_by_order_id': agg_func,\n",
    "        'order_ts' : agg_func,\n",
    "        'order_ts_diff' : agg_func,\n",
    "        'quantity_diff' : agg_func,\n",
    "        'price_diff' : agg_func,\n",
    "        'total_diff' : agg_func,\n",
    "        'order_id': ['nunique'],\n",
    "        'product_id': ['nunique'],\n",
    "    }\n",
    "    all_train_data = pd.DataFrame()\n",
    "    \n",
    "    for i, tr_ym in enumerate(train_label['year_month'].unique()):\n",
    "        # group by aggretation 함수로 train 데이터 피처 생성\n",
    "        train_agg = train.loc[train['order_date'] < tr_ym].groupby(['customer_id']).agg(agg_dict)\n",
    "                \n",
    "        # 멀티 레벨 컬럼을 사용하기 쉽게 1 레벨 컬럼명으로 변경\n",
    "        new_cols = []\n",
    "        for col in agg_dict.keys():\n",
    "            for stat in agg_dict[col]:\n",
    "                if type(stat) is str:\n",
    "                    new_cols.append(f'{col}-{stat}')\n",
    "                else:\n",
    "                    new_cols.append(f'{col}-mode')\n",
    "\n",
    "#         for col in train_agg.columns.levels[0]:\n",
    "#             for stat in train_agg.columns.levels[1]:\n",
    "#                 new_cols.append(f'{col}-{stat}')\n",
    "        \n",
    "        train_agg.columns = new_cols\n",
    "        train_agg.reset_index(inplace = True)\n",
    "        \n",
    "        train_agg['year_month'] = tr_ym\n",
    "        \n",
    "        all_train_data = all_train_data.append(train_agg)\n",
    "        \n",
    "    \n",
    "    all_train_data = train_label.merge(all_train_data, on=['customer_id', 'year_month'], how='left')\n",
    "    \n",
    "    features = all_train_data.drop(columns=['customer_id', 'label', 'year_month']).columns\n",
    "    \n",
    "    # group by aggretation 함수로 test 데이터 피처 생성\n",
    "    test_agg = test.groupby(['customer_id']).agg(agg_dict)\n",
    "    test_agg.columns = new_cols\n",
    "    \n",
    "    \n",
    "    test_data = test_label.merge(test_agg, on=['customer_id'], how='left')\n",
    "    # train, test 데이터 전처리\n",
    "    x_tr, x_te = feature_preprocessing1(all_train_data, test_data, features)\n",
    "    \n",
    "    print('x_tr.shape', x_tr.shape, ', x_te.shape', x_te.shape)\n",
    "    \n",
    "    return x_tr, x_te, all_train_data['label'], features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "id": "928f6016",
   "metadata": {},
   "outputs": [],
   "source": [
    "def make_lgb_oof_prediction(train, y, test, features, categorical_features='auto', model_params=None, folds=10):\n",
    "    x_train = train[features]\n",
    "    x_test = test[features]\n",
    "    \n",
    "    # 테스트 데이터 예측값을 저장할 변수\n",
    "    test_preds = np.zeros(x_test.shape[0])\n",
    "    \n",
    "    # Out Of Fold Validation 예측 데이터를 저장할 변수\n",
    "    y_oof = np.zeros(x_train.shape[0])\n",
    "    \n",
    "    # 폴드별 평균 Validation 스코어를 저장할 변수\n",
    "    score = 0\n",
    "    \n",
    "    # 피처 중요도를 저장할 데이터 프레임 선언\n",
    "    fi = pd.DataFrame()\n",
    "    fi['feature'] = features\n",
    "    \n",
    "    # Stratified K Fold 선언\n",
    "    skf = StratifiedKFold(n_splits=folds, shuffle=True, random_state=SEED)\n",
    "\n",
    "    for fold, (tr_idx, val_idx) in enumerate(skf.split(x_train, y)):\n",
    "        # train index, validation index로 train 데이터를 나눔\n",
    "        x_tr, x_val = x_train.loc[tr_idx, features], x_train.loc[val_idx, features]\n",
    "        y_tr, y_val = y[tr_idx], y[val_idx]\n",
    "        \n",
    "        print(f'fold: {fold+1}, x_tr.shape: {x_tr.shape}, x_val.shape: {x_val.shape}')\n",
    "\n",
    "        # LightGBM 데이터셋 선언\n",
    "        dtrain = lgb.Dataset(x_tr, label=y_tr)\n",
    "        dvalid = lgb.Dataset(x_val, label=y_val)\n",
    "        \n",
    "        # LightGBM 모델 훈련\n",
    "        clf = lgb.train(\n",
    "            model_params,\n",
    "            dtrain,\n",
    "            valid_sets=[dtrain, dvalid], # Validation 성능을 측정할 수 있도록 설정\n",
    "            categorical_feature=categorical_features,\n",
    "            verbose_eval=200\n",
    "        )\n",
    "\n",
    "        # Validation 데이터 예측\n",
    "        val_preds = clf.predict(x_val)\n",
    "        \n",
    "        # Validation index에 예측값 저장 \n",
    "        y_oof[val_idx] = val_preds\n",
    "        \n",
    "        # 폴드별 Validation 스코어 측정\n",
    "        print(f\"Fold {fold + 1} | AUC: {roc_auc_score(y_val, val_preds)}\")\n",
    "        print('-'*80)\n",
    "\n",
    "        # score 변수에 폴드별 평균 Validation 스코어 저장\n",
    "        score += roc_auc_score(y_val, val_preds) / folds\n",
    "        \n",
    "        # 테스트 데이터 예측하고 평균해서 저장\n",
    "        test_preds += clf.predict(x_test) / folds\n",
    "        \n",
    "        # 폴드별 피처 중요도 저장\n",
    "        fi[f'fold_{fold+1}'] = clf.feature_importance()\n",
    "\n",
    "        del x_tr, x_val, y_tr, y_val\n",
    "        gc.collect()\n",
    "        \n",
    "    print(f\"\\nMean AUC = {score}\") # 폴드별 Validation 스코어 출력\n",
    "    print(f\"OOF AUC = {roc_auc_score(y, y_oof)}\") # Out Of Fold Validation 스코어 출력\n",
    "        \n",
    "    # 폴드별 피처 중요도 평균값 계산해서 저장 \n",
    "    fi_cols = [col for col in fi.columns if 'fold_' in col]\n",
    "    fi['importance'] = fi[fi_cols].mean(axis=1)\n",
    "    \n",
    "    return y_oof, test_preds, fi"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "id": "d76098a9",
   "metadata": {},
   "outputs": [],
   "source": [
    "def make_cat_oof_prediction(train, y, test, features, categorical_features=None, model_params=None, folds = 10):\n",
    "    x_train = train[features]\n",
    "    x_test = test[features]\n",
    "    \n",
    "    test_preds = np.zeros(x_test.shape[0])\n",
    "    \n",
    "    y_oof = np.zeros(x_train.shape[0])\n",
    "    \n",
    "    score = 0\n",
    "    \n",
    "    fi = pd.DataFrame()\n",
    "    fi['feature'] = features\n",
    "    \n",
    "    skf = StratifiedKFold(n_splits=folds, shuffle=True, random_state=SEED)\n",
    "    \n",
    "    for fold, (tr_idx, val_idx) in enumerate(skf.split(x_train,y)):\n",
    "        x_tr, x_val = x_train.loc[tr_idx, features], x_train.loc[val_idx, features]\n",
    "        y_tr, y_val = y[tr_idx], y[val_idx]\n",
    "        \n",
    "        print(f'fold: {fold+1}, x_tr.shape: {x_tr.shape}, x_val.shape: {x_val.shape}')\n",
    "        \n",
    "        clf = CatBoostClassifier(**model_params)\n",
    "        clf.fit(x_tr, y_tr,\n",
    "                eval_set=(x_val, y_val),\n",
    "                cat_features=categorical_features,\n",
    "                use_best_model=True,\n",
    "                verbose=True)\n",
    "        \n",
    "        val_preds = clf.predict_proba(x_val)[:,1]\n",
    "        \n",
    "        y_oof[val_idx] = val_preds\n",
    "        \n",
    "        print(f'Fold {fold+1} | AUC:{roc_auc_score(y_val, val_preds)}')\n",
    "        print('-'*80)\n",
    "        \n",
    "        score += roc_auc_score(y_val, val_preds) / folds\n",
    "        \n",
    "        test_preds += clf.predict_proba(x_test)[:,1] / folds\n",
    "        \n",
    "        fi[f'fold_{fold+1}'] = clf.feature_importances_\n",
    "        \n",
    "        del x_tr, x_val, y_tr, y_val\n",
    "        gc.collect()\n",
    "        \n",
    "    print(f'\\nMean AUC = {score}')\n",
    "    print(f'OOF AUC = {roc_auc_score(y, y_oof)}')\n",
    "    \n",
    "    fi_cols = [col for col in fi.columns if 'fold_' in col]\n",
    "    fi['importance'] = fi[fi_cols].mean(axis = 1)\n",
    "    \n",
    "    return y_oof, test_preds, fi\n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "id": "9a81f223",
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_feature_importances(df, n=20, color='blue', figsize=(12,8)):\n",
    "    # 피처 중요도 순으로 내림차순 정렬\n",
    "    df = df.sort_values('importance', ascending = False).reset_index(drop = True)\n",
    "    \n",
    "    # 피처 중요도 정규화 및 누적 중요도 계산\n",
    "    df['importance_normalized'] = df['importance'] / df['importance'].sum()\n",
    "    df['cumulative_importance'] = np.cumsum(df['importance_normalized'])\n",
    "    \n",
    "    plt.rcParams['font.size'] = 12\n",
    "    plt.style.use('fivethirtyeight')\n",
    "    # 피처 중요도 순으로 n개까지 바플롯으로 그리기\n",
    "    df.loc[:n, :].plot.barh(y='importance_normalized', \n",
    "                            x='feature', color=color, \n",
    "                            edgecolor='k', figsize=figsize,\n",
    "                            legend=False)\n",
    "\n",
    "    plt.xlabel('Normalized Importance', size=18); plt.ylabel(''); \n",
    "    plt.title(f'Top {n} Most Important Features', size=18)\n",
    "    plt.gca().invert_yaxis()\n",
    "    \n",
    "    return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "id": "88a35a8f",
   "metadata": {},
   "outputs": [],
   "source": [
    "lgb_params = {\n",
    "    'objective': 'binary', # 이진 분류\n",
    "    'boosting_type': 'gbdt',\n",
    "    'metric': 'auc', # 평가 지표 설정\n",
    "    'feature_fraction': 0.8, # 피처 샘플링 비율\n",
    "    'bagging_fraction': 0.8, # 데이터 샘플링 비율\n",
    "    'bagging_freq': 1,\n",
    "    'n_estimators': 10000, # 트리 개수\n",
    "    'early_stopping_rounds': 100,\n",
    "    'seed': SEED,\n",
    "    'verbose': -1,\n",
    "    'n_jobs': -1,    \n",
    "}\n",
    "\n",
    "tuned_lgb_params = {\n",
    "    'objective': 'binary', # 이진 분류\n",
    "    'boosting_type': 'gbdt',\n",
    "    'metric': 'auc', # 평가 지표 설정\n",
    "    'feature_fraction': 0.9068781382922049, # 피처 샘플링 비율\n",
    "    'bagging_fraction': 0.5427572227580095, # 데이터 샘플링 비율\n",
    "    'bagging_freq': 4,\n",
    "    'n_estimators': 10000, # 트리 개수\n",
    "    'early_stopping_rounds': 100,\n",
    "    'seed': SEED,\n",
    "    'verbose': -1,\n",
    "    'n_jobs': -1, \n",
    "    'num_leaves': 159,\n",
    "    'max_bin': 135,\n",
    "    'min_data_in_leaf': 28,\n",
    "    'lambda_l1': 2.531979160239369,\n",
    "    'lambda_l2': 1.645404534997536e-05,\n",
    "}\n",
    "\n",
    "cat_params = {\n",
    "    'n_estimators': 10000, # 트리 개수\n",
    "    'learning_rate': 0.07, # 학습률\n",
    "    'eval_metric': 'AUC', # 평가 지표 설정\n",
    "    'loss_function': 'Logloss', # 손실 함수 설정\n",
    "    'random_seed': SEED,\n",
    "    'metric_period': 100,\n",
    "    'od_wait': 100, # early stopping round\n",
    "    'depth': 6, # 트리 최고 깊이\n",
    "    'rsm': 0.8, # 피처 샘플링 비율\n",
    "}\n",
    "\n",
    "tuned_cat_params = {\n",
    "    #'n_estimators': 10000, # 트리 개수\n",
    "    'learning_rate': 0.06173991466879898, # 학습률\n",
    "    'eval_metric': 'AUC', # 평가 지표 설정\n",
    "    'loss_function': 'Logloss', # 손실 함수 설정\n",
    "    'random_seed': SEED,\n",
    "    'metric_period': 100,\n",
    "    'od_wait': 100, # early stopping round\n",
    "    'depth': 5, # 트리 최고 깊이\n",
    "    'rsm': 0.8, # 피처 샘플링 비율\n",
    "    'iterations': 180,\n",
    "    'random_strength': 65,\n",
    "    'bagging_temperature': 0.43432569253026015,\n",
    "    'od_type': 'Iter'\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "id": "e795b152",
   "metadata": {},
   "outputs": [],
   "source": [
    "label_2011_11 = generate_label(df, '2011-11')['label']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "id": "59c3a553",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "categorical feature: []\n",
      "x_tr.shape (5501, 124) , x_te.shape (5722, 124)\n"
     ]
    }
   ],
   "source": [
    "train, test, y, features = feature_engineering1(df, '2011-11')\n",
    "#y_oof, test_preds_2011_11, fi = make_lgb_oof_prediction(train, y, test, features, model_params=lgb_params)\n",
    "#print_score(label_2011_11, test_preds_2011_11)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "6ea30981",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>customer_id</th>\n",
       "      <th>year_month</th>\n",
       "      <th>label</th>\n",
       "      <th>cumsum_total_by_cust_id-mean</th>\n",
       "      <th>cumsum_total_by_cust_id-max</th>\n",
       "      <th>cumsum_total_by_cust_id-min</th>\n",
       "      <th>cumsum_total_by_cust_id-sum</th>\n",
       "      <th>cumsum_total_by_cust_id-count</th>\n",
       "      <th>cumsum_total_by_cust_id-std</th>\n",
       "      <th>cumsum_total_by_cust_id-skew</th>\n",
       "      <th>cumsum_quantity_by_cust_id-mean</th>\n",
       "      <th>cumsum_quantity_by_cust_id-max</th>\n",
       "      <th>cumsum_quantity_by_cust_id-min</th>\n",
       "      <th>cumsum_quantity_by_cust_id-sum</th>\n",
       "      <th>cumsum_quantity_by_cust_id-count</th>\n",
       "      <th>cumsum_quantity_by_cust_id-std</th>\n",
       "      <th>cumsum_quantity_by_cust_id-skew</th>\n",
       "      <th>cumsum_price_by_cust_id-mean</th>\n",
       "      <th>cumsum_price_by_cust_id-max</th>\n",
       "      <th>cumsum_price_by_cust_id-min</th>\n",
       "      <th>cumsum_price_by_cust_id-sum</th>\n",
       "      <th>cumsum_price_by_cust_id-count</th>\n",
       "      <th>cumsum_price_by_cust_id-std</th>\n",
       "      <th>cumsum_price_by_cust_id-skew</th>\n",
       "      <th>cumsum_total_by_prod_id-mean</th>\n",
       "      <th>cumsum_total_by_prod_id-max</th>\n",
       "      <th>cumsum_total_by_prod_id-min</th>\n",
       "      <th>cumsum_total_by_prod_id-sum</th>\n",
       "      <th>cumsum_total_by_prod_id-count</th>\n",
       "      <th>cumsum_total_by_prod_id-std</th>\n",
       "      <th>cumsum_total_by_prod_id-skew</th>\n",
       "      <th>cumsum_quantity_by_prod_id-mean</th>\n",
       "      <th>cumsum_quantity_by_prod_id-max</th>\n",
       "      <th>cumsum_quantity_by_prod_id-min</th>\n",
       "      <th>cumsum_quantity_by_prod_id-sum</th>\n",
       "      <th>cumsum_quantity_by_prod_id-count</th>\n",
       "      <th>cumsum_quantity_by_prod_id-std</th>\n",
       "      <th>cumsum_quantity_by_prod_id-skew</th>\n",
       "      <th>cumsum_price_by_prod_id-mean</th>\n",
       "      <th>cumsum_price_by_prod_id-max</th>\n",
       "      <th>cumsum_price_by_prod_id-min</th>\n",
       "      <th>cumsum_price_by_prod_id-sum</th>\n",
       "      <th>cumsum_price_by_prod_id-count</th>\n",
       "      <th>cumsum_price_by_prod_id-std</th>\n",
       "      <th>cumsum_price_by_prod_id-skew</th>\n",
       "      <th>cumsum_total_by_order_id-mean</th>\n",
       "      <th>cumsum_total_by_order_id-max</th>\n",
       "      <th>cumsum_total_by_order_id-min</th>\n",
       "      <th>cumsum_total_by_order_id-sum</th>\n",
       "      <th>cumsum_total_by_order_id-count</th>\n",
       "      <th>cumsum_total_by_order_id-std</th>\n",
       "      <th>cumsum_total_by_order_id-skew</th>\n",
       "      <th>cumsum_quantity_by_order_id-mean</th>\n",
       "      <th>cumsum_quantity_by_order_id-max</th>\n",
       "      <th>cumsum_quantity_by_order_id-min</th>\n",
       "      <th>cumsum_quantity_by_order_id-sum</th>\n",
       "      <th>cumsum_quantity_by_order_id-count</th>\n",
       "      <th>cumsum_quantity_by_order_id-std</th>\n",
       "      <th>cumsum_quantity_by_order_id-skew</th>\n",
       "      <th>cumsum_price_by_order_id-mean</th>\n",
       "      <th>cumsum_price_by_order_id-max</th>\n",
       "      <th>cumsum_price_by_order_id-min</th>\n",
       "      <th>cumsum_price_by_order_id-sum</th>\n",
       "      <th>cumsum_price_by_order_id-count</th>\n",
       "      <th>cumsum_price_by_order_id-std</th>\n",
       "      <th>cumsum_price_by_order_id-skew</th>\n",
       "      <th>order_ts-mean</th>\n",
       "      <th>order_ts-max</th>\n",
       "      <th>order_ts-min</th>\n",
       "      <th>order_ts-sum</th>\n",
       "      <th>order_ts-count</th>\n",
       "      <th>order_ts-std</th>\n",
       "      <th>order_ts-skew</th>\n",
       "      <th>order_ts_diff-mean</th>\n",
       "      <th>order_ts_diff-max</th>\n",
       "      <th>order_ts_diff-min</th>\n",
       "      <th>order_ts_diff-sum</th>\n",
       "      <th>order_ts_diff-count</th>\n",
       "      <th>order_ts_diff-std</th>\n",
       "      <th>order_ts_diff-skew</th>\n",
       "      <th>quantity_diff-mean</th>\n",
       "      <th>quantity_diff-max</th>\n",
       "      <th>quantity_diff-min</th>\n",
       "      <th>quantity_diff-sum</th>\n",
       "      <th>quantity_diff-count</th>\n",
       "      <th>quantity_diff-std</th>\n",
       "      <th>quantity_diff-skew</th>\n",
       "      <th>price_diff-mean</th>\n",
       "      <th>price_diff-max</th>\n",
       "      <th>price_diff-min</th>\n",
       "      <th>price_diff-sum</th>\n",
       "      <th>price_diff-count</th>\n",
       "      <th>price_diff-std</th>\n",
       "      <th>price_diff-skew</th>\n",
       "      <th>total_diff-mean</th>\n",
       "      <th>total_diff-max</th>\n",
       "      <th>total_diff-min</th>\n",
       "      <th>total_diff-sum</th>\n",
       "      <th>total_diff-count</th>\n",
       "      <th>total_diff-std</th>\n",
       "      <th>total_diff-skew</th>\n",
       "      <th>order_id-nunique</th>\n",
       "      <th>product_id-nunique</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>12346</td>\n",
       "      <td>2011-10</td>\n",
       "      <td>0</td>\n",
       "      <td>2953.304457</td>\n",
       "      <td>127267.5690</td>\n",
       "      <td>-85.3710</td>\n",
       "      <td>138805.3095</td>\n",
       "      <td>47.0</td>\n",
       "      <td>18527.811314</td>\n",
       "      <td>6.855058</td>\n",
       "      <td>1631.063830</td>\n",
       "      <td>74268.0</td>\n",
       "      <td>10.0</td>\n",
       "      <td>76660.0</td>\n",
       "      <td>47.0</td>\n",
       "      <td>10825.525671</td>\n",
       "      <td>6.855636</td>\n",
       "      <td>400.641415</td>\n",
       "      <td>936.3750</td>\n",
       "      <td>7.4250</td>\n",
       "      <td>18830.1465</td>\n",
       "      <td>47.0</td>\n",
       "      <td>262.090504</td>\n",
       "      <td>0.440160</td>\n",
       "      <td>-18933.867255</td>\n",
       "      <td>127352.9400</td>\n",
       "      <td>-145015.2000</td>\n",
       "      <td>-8.898918e+05</td>\n",
       "      <td>47.0</td>\n",
       "      <td>64968.445670</td>\n",
       "      <td>-1.131166</td>\n",
       "      <td>2354.212766</td>\n",
       "      <td>74215.0</td>\n",
       "      <td>-1644.0</td>\n",
       "      <td>110648.0</td>\n",
       "      <td>47.0</td>\n",
       "      <td>10767.598390</td>\n",
       "      <td>6.742285</td>\n",
       "      <td>63650.017691</td>\n",
       "      <td>325951.2630</td>\n",
       "      <td>1.6500</td>\n",
       "      <td>2.991551e+06</td>\n",
       "      <td>47.0</td>\n",
       "      <td>128903.599171</td>\n",
       "      <td>1.619409</td>\n",
       "      <td>24.297128</td>\n",
       "      <td>127352.9400</td>\n",
       "      <td>-127352.9400</td>\n",
       "      <td>1141.9650</td>\n",
       "      <td>47.0</td>\n",
       "      <td>26555.274233</td>\n",
       "      <td>-0.002867</td>\n",
       "      <td>3.914894</td>\n",
       "      <td>74215.0</td>\n",
       "      <td>-74215.0</td>\n",
       "      <td>184.0</td>\n",
       "      <td>47.0</td>\n",
       "      <td>15474.899161</td>\n",
       "      <td>-0.000793</td>\n",
       "      <td>80.737660</td>\n",
       "      <td>412.5990</td>\n",
       "      <td>1.6500</td>\n",
       "      <td>3794.6700</td>\n",
       "      <td>47.0</td>\n",
       "      <td>87.485676</td>\n",
       "      <td>1.486909</td>\n",
       "      <td>1.274484e+09</td>\n",
       "      <td>1.295346e+09</td>\n",
       "      <td>1.260780e+09</td>\n",
       "      <td>5.990076e+10</td>\n",
       "      <td>47.0</td>\n",
       "      <td>8.367102e+06</td>\n",
       "      <td>0.056381</td>\n",
       "      <td>751438.695652</td>\n",
       "      <td>10197900.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>34566180.0</td>\n",
       "      <td>46.0</td>\n",
       "      <td>2.323281e+06</td>\n",
       "      <td>3.439852</td>\n",
       "      <td>-1613.586957</td>\n",
       "      <td>74216.0</td>\n",
       "      <td>-148430.0</td>\n",
       "      <td>-74225.0</td>\n",
       "      <td>46.0</td>\n",
       "      <td>24684.548109</td>\n",
       "      <td>-4.214331</td>\n",
       "      <td>-1.241087e-01</td>\n",
       "      <td>182.3745</td>\n",
       "      <td>-233.1615</td>\n",
       "      <td>-5.709000e+00</td>\n",
       "      <td>46.0</td>\n",
       "      <td>56.879462</td>\n",
       "      <td>-0.840908</td>\n",
       "      <td>-2770.156304</td>\n",
       "      <td>127360.365</td>\n",
       "      <td>-254705.880</td>\n",
       "      <td>-127427.1900</td>\n",
       "      <td>46.0</td>\n",
       "      <td>42359.030338</td>\n",
       "      <td>-4.214047</td>\n",
       "      <td>17.0</td>\n",
       "      <td>30.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>12347</td>\n",
       "      <td>2011-10</td>\n",
       "      <td>1</td>\n",
       "      <td>2546.704976</td>\n",
       "      <td>5613.9435</td>\n",
       "      <td>29.2050</td>\n",
       "      <td>417659.6160</td>\n",
       "      <td>164.0</td>\n",
       "      <td>1636.394946</td>\n",
       "      <td>0.120206</td>\n",
       "      <td>1031.786585</td>\n",
       "      <td>2099.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>169213.0</td>\n",
       "      <td>164.0</td>\n",
       "      <td>606.713512</td>\n",
       "      <td>0.118385</td>\n",
       "      <td>317.244293</td>\n",
       "      <td>693.4455</td>\n",
       "      <td>4.8675</td>\n",
       "      <td>52028.0640</td>\n",
       "      <td>164.0</td>\n",
       "      <td>207.325091</td>\n",
       "      <td>0.111363</td>\n",
       "      <td>14046.581799</td>\n",
       "      <td>363807.0975</td>\n",
       "      <td>156.3375</td>\n",
       "      <td>2.303639e+06</td>\n",
       "      <td>164.0</td>\n",
       "      <td>43323.609176</td>\n",
       "      <td>6.372546</td>\n",
       "      <td>4248.969512</td>\n",
       "      <td>45875.0</td>\n",
       "      <td>56.0</td>\n",
       "      <td>696831.0</td>\n",
       "      <td>164.0</td>\n",
       "      <td>7357.646380</td>\n",
       "      <td>3.028850</td>\n",
       "      <td>1940.714909</td>\n",
       "      <td>61460.1900</td>\n",
       "      <td>8.2500</td>\n",
       "      <td>3.182772e+05</td>\n",
       "      <td>164.0</td>\n",
       "      <td>7252.527668</td>\n",
       "      <td>6.633689</td>\n",
       "      <td>470.958796</td>\n",
       "      <td>1174.4535</td>\n",
       "      <td>21.7800</td>\n",
       "      <td>77237.2425</td>\n",
       "      <td>164.0</td>\n",
       "      <td>291.682227</td>\n",
       "      <td>0.353722</td>\n",
       "      <td>199.012195</td>\n",
       "      <td>509.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>32638.0</td>\n",
       "      <td>164.0</td>\n",
       "      <td>134.237770</td>\n",
       "      <td>0.480564</td>\n",
       "      <td>57.845881</td>\n",
       "      <td>147.8235</td>\n",
       "      <td>0.9075</td>\n",
       "      <td>9486.7245</td>\n",
       "      <td>164.0</td>\n",
       "      <td>35.377236</td>\n",
       "      <td>0.300351</td>\n",
       "      <td>1.297744e+09</td>\n",
       "      <td>1.312275e+09</td>\n",
       "      <td>1.288535e+09</td>\n",
       "      <td>2.128301e+11</td>\n",
       "      <td>164.0</td>\n",
       "      <td>8.387982e+06</td>\n",
       "      <td>0.513752</td>\n",
       "      <td>145644.662577</td>\n",
       "      <td>6120780.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>23740080.0</td>\n",
       "      <td>163.0</td>\n",
       "      <td>8.397292e+05</td>\n",
       "      <td>5.854925</td>\n",
       "      <td>0.036810</td>\n",
       "      <td>234.0</td>\n",
       "      <td>-228.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>163.0</td>\n",
       "      <td>27.372576</td>\n",
       "      <td>0.288404</td>\n",
       "      <td>5.263804e-02</td>\n",
       "      <td>20.1300</td>\n",
       "      <td>-16.1700</td>\n",
       "      <td>8.580000e+00</td>\n",
       "      <td>163.0</td>\n",
       "      <td>4.747310</td>\n",
       "      <td>0.693660</td>\n",
       "      <td>0.810828</td>\n",
       "      <td>382.635</td>\n",
       "      <td>-387.090</td>\n",
       "      <td>132.1650</td>\n",
       "      <td>163.0</td>\n",
       "      <td>50.079192</td>\n",
       "      <td>0.000551</td>\n",
       "      <td>6.0</td>\n",
       "      <td>107.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>12348</td>\n",
       "      <td>2011-10</td>\n",
       "      <td>0</td>\n",
       "      <td>1096.986000</td>\n",
       "      <td>3332.0100</td>\n",
       "      <td>21.7800</td>\n",
       "      <td>55946.2860</td>\n",
       "      <td>51.0</td>\n",
       "      <td>1000.019165</td>\n",
       "      <td>0.812596</td>\n",
       "      <td>1069.450980</td>\n",
       "      <td>2714.0</td>\n",
       "      <td>24.0</td>\n",
       "      <td>54542.0</td>\n",
       "      <td>51.0</td>\n",
       "      <td>902.451734</td>\n",
       "      <td>0.493378</td>\n",
       "      <td>66.215794</td>\n",
       "      <td>318.6150</td>\n",
       "      <td>0.9075</td>\n",
       "      <td>3377.0055</td>\n",
       "      <td>51.0</td>\n",
       "      <td>79.524917</td>\n",
       "      <td>1.597336</td>\n",
       "      <td>17500.335471</td>\n",
       "      <td>146857.4085</td>\n",
       "      <td>471.2400</td>\n",
       "      <td>8.925171e+05</td>\n",
       "      <td>51.0</td>\n",
       "      <td>27894.026806</td>\n",
       "      <td>3.120231</td>\n",
       "      <td>11485.372549</td>\n",
       "      <td>35932.0</td>\n",
       "      <td>336.0</td>\n",
       "      <td>585754.0</td>\n",
       "      <td>51.0</td>\n",
       "      <td>11056.078952</td>\n",
       "      <td>0.673806</td>\n",
       "      <td>5104.847000</td>\n",
       "      <td>95332.3635</td>\n",
       "      <td>25.1955</td>\n",
       "      <td>2.603472e+05</td>\n",
       "      <td>51.0</td>\n",
       "      <td>16507.639799</td>\n",
       "      <td>4.106796</td>\n",
       "      <td>375.299294</td>\n",
       "      <td>1473.1200</td>\n",
       "      <td>21.7800</td>\n",
       "      <td>19140.2640</td>\n",
       "      <td>51.0</td>\n",
       "      <td>300.618991</td>\n",
       "      <td>1.639190</td>\n",
       "      <td>388.392157</td>\n",
       "      <td>1254.0</td>\n",
       "      <td>24.0</td>\n",
       "      <td>19808.0</td>\n",
       "      <td>51.0</td>\n",
       "      <td>333.951917</td>\n",
       "      <td>1.339273</td>\n",
       "      <td>14.115588</td>\n",
       "      <td>81.8400</td>\n",
       "      <td>0.4785</td>\n",
       "      <td>719.8950</td>\n",
       "      <td>51.0</td>\n",
       "      <td>18.844181</td>\n",
       "      <td>2.593528</td>\n",
       "      <td>1.292579e+09</td>\n",
       "      <td>1.316956e+09</td>\n",
       "      <td>1.285600e+09</td>\n",
       "      <td>6.592153e+10</td>\n",
       "      <td>51.0</td>\n",
       "      <td>8.085985e+06</td>\n",
       "      <td>1.617142</td>\n",
       "      <td>627136.800000</td>\n",
       "      <td>14955960.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>31356840.0</td>\n",
       "      <td>50.0</td>\n",
       "      <td>2.475659e+06</td>\n",
       "      <td>4.710186</td>\n",
       "      <td>-0.460000</td>\n",
       "      <td>138.0</td>\n",
       "      <td>-143.0</td>\n",
       "      <td>-23.0</td>\n",
       "      <td>50.0</td>\n",
       "      <td>63.018692</td>\n",
       "      <td>0.066540</td>\n",
       "      <td>1.301850e+00</td>\n",
       "      <td>65.5215</td>\n",
       "      <td>-65.5215</td>\n",
       "      <td>6.509250e+01</td>\n",
       "      <td>50.0</td>\n",
       "      <td>24.419082</td>\n",
       "      <td>0.251831</td>\n",
       "      <td>0.884400</td>\n",
       "      <td>312.840</td>\n",
       "      <td>-327.096</td>\n",
       "      <td>44.2200</td>\n",
       "      <td>50.0</td>\n",
       "      <td>83.855397</td>\n",
       "      <td>-0.069972</td>\n",
       "      <td>5.0</td>\n",
       "      <td>25.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>12349</td>\n",
       "      <td>2011-10</td>\n",
       "      <td>0</td>\n",
       "      <td>2014.247369</td>\n",
       "      <td>4367.5335</td>\n",
       "      <td>-39.8475</td>\n",
       "      <td>215524.4685</td>\n",
       "      <td>107.0</td>\n",
       "      <td>1402.866250</td>\n",
       "      <td>0.089354</td>\n",
       "      <td>467.682243</td>\n",
       "      <td>988.0</td>\n",
       "      <td>-5.0</td>\n",
       "      <td>50042.0</td>\n",
       "      <td>107.0</td>\n",
       "      <td>273.727436</td>\n",
       "      <td>-0.077728</td>\n",
       "      <td>615.014523</td>\n",
       "      <td>1484.1585</td>\n",
       "      <td>6.1875</td>\n",
       "      <td>65806.5540</td>\n",
       "      <td>107.0</td>\n",
       "      <td>484.322565</td>\n",
       "      <td>0.447134</td>\n",
       "      <td>8876.528131</td>\n",
       "      <td>86587.3965</td>\n",
       "      <td>74.2500</td>\n",
       "      <td>9.497885e+05</td>\n",
       "      <td>107.0</td>\n",
       "      <td>12932.316422</td>\n",
       "      <td>3.313619</td>\n",
       "      <td>2352.869159</td>\n",
       "      <td>33089.0</td>\n",
       "      <td>12.0</td>\n",
       "      <td>251757.0</td>\n",
       "      <td>107.0</td>\n",
       "      <td>4244.373963</td>\n",
       "      <td>4.872195</td>\n",
       "      <td>1555.892636</td>\n",
       "      <td>30169.8210</td>\n",
       "      <td>24.2550</td>\n",
       "      <td>1.664805e+05</td>\n",
       "      <td>107.0</td>\n",
       "      <td>3418.294904</td>\n",
       "      <td>6.236526</td>\n",
       "      <td>959.884724</td>\n",
       "      <td>2314.3230</td>\n",
       "      <td>-39.8475</td>\n",
       "      <td>102707.6655</td>\n",
       "      <td>107.0</td>\n",
       "      <td>703.447214</td>\n",
       "      <td>0.321784</td>\n",
       "      <td>224.383178</td>\n",
       "      <td>519.0</td>\n",
       "      <td>-5.0</td>\n",
       "      <td>24009.0</td>\n",
       "      <td>107.0</td>\n",
       "      <td>135.943082</td>\n",
       "      <td>0.244650</td>\n",
       "      <td>239.373056</td>\n",
       "      <td>793.2540</td>\n",
       "      <td>2.7225</td>\n",
       "      <td>25612.9170</td>\n",
       "      <td>107.0</td>\n",
       "      <td>248.595111</td>\n",
       "      <td>1.326011</td>\n",
       "      <td>1.280047e+09</td>\n",
       "      <td>1.288254e+09</td>\n",
       "      <td>1.259931e+09</td>\n",
       "      <td>1.369650e+11</td>\n",
       "      <td>107.0</td>\n",
       "      <td>8.875333e+06</td>\n",
       "      <td>-0.411039</td>\n",
       "      <td>267200.377358</td>\n",
       "      <td>14077560.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>28323240.0</td>\n",
       "      <td>106.0</td>\n",
       "      <td>1.832081e+06</td>\n",
       "      <td>7.131391</td>\n",
       "      <td>0.103774</td>\n",
       "      <td>36.0</td>\n",
       "      <td>-36.0</td>\n",
       "      <td>11.0</td>\n",
       "      <td>106.0</td>\n",
       "      <td>8.632088</td>\n",
       "      <td>-0.091287</td>\n",
       "      <td>-3.268868e-02</td>\n",
       "      <td>399.3825</td>\n",
       "      <td>-411.0975</td>\n",
       "      <td>-3.465000e+00</td>\n",
       "      <td>106.0</td>\n",
       "      <td>71.657823</td>\n",
       "      <td>-0.213999</td>\n",
       "      <td>0.315212</td>\n",
       "      <td>386.265</td>\n",
       "      <td>-395.670</td>\n",
       "      <td>33.4125</td>\n",
       "      <td>106.0</td>\n",
       "      <td>75.363508</td>\n",
       "      <td>-0.089535</td>\n",
       "      <td>4.0</td>\n",
       "      <td>92.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>12350</td>\n",
       "      <td>2011-10</td>\n",
       "      <td>0</td>\n",
       "      <td>281.849118</td>\n",
       "      <td>551.7600</td>\n",
       "      <td>41.5800</td>\n",
       "      <td>4791.4350</td>\n",
       "      <td>17.0</td>\n",
       "      <td>161.979220</td>\n",
       "      <td>0.200043</td>\n",
       "      <td>107.941176</td>\n",
       "      <td>197.0</td>\n",
       "      <td>12.0</td>\n",
       "      <td>1835.0</td>\n",
       "      <td>17.0</td>\n",
       "      <td>57.957388</td>\n",
       "      <td>-0.168685</td>\n",
       "      <td>42.133235</td>\n",
       "      <td>107.7450</td>\n",
       "      <td>3.4650</td>\n",
       "      <td>716.2650</td>\n",
       "      <td>17.0</td>\n",
       "      <td>40.745863</td>\n",
       "      <td>0.879343</td>\n",
       "      <td>12878.018029</td>\n",
       "      <td>88027.8630</td>\n",
       "      <td>292.0500</td>\n",
       "      <td>2.189263e+05</td>\n",
       "      <td>17.0</td>\n",
       "      <td>21329.326160</td>\n",
       "      <td>3.050607</td>\n",
       "      <td>3653.352941</td>\n",
       "      <td>16276.0</td>\n",
       "      <td>60.0</td>\n",
       "      <td>62107.0</td>\n",
       "      <td>17.0</td>\n",
       "      <td>4338.504177</td>\n",
       "      <td>1.810547</td>\n",
       "      <td>3051.218824</td>\n",
       "      <td>41075.8260</td>\n",
       "      <td>77.8800</td>\n",
       "      <td>5.187072e+04</td>\n",
       "      <td>17.0</td>\n",
       "      <td>9814.784063</td>\n",
       "      <td>4.100398</td>\n",
       "      <td>281.849118</td>\n",
       "      <td>551.7600</td>\n",
       "      <td>41.5800</td>\n",
       "      <td>4791.4350</td>\n",
       "      <td>17.0</td>\n",
       "      <td>161.979220</td>\n",
       "      <td>0.200043</td>\n",
       "      <td>107.941176</td>\n",
       "      <td>197.0</td>\n",
       "      <td>12.0</td>\n",
       "      <td>1835.0</td>\n",
       "      <td>17.0</td>\n",
       "      <td>57.957388</td>\n",
       "      <td>-0.168685</td>\n",
       "      <td>42.133235</td>\n",
       "      <td>107.7450</td>\n",
       "      <td>3.4650</td>\n",
       "      <td>716.2650</td>\n",
       "      <td>17.0</td>\n",
       "      <td>40.745863</td>\n",
       "      <td>0.879343</td>\n",
       "      <td>1.296662e+09</td>\n",
       "      <td>1.296662e+09</td>\n",
       "      <td>1.296662e+09</td>\n",
       "      <td>2.204326e+10</td>\n",
       "      <td>17.0</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>16.0</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>12.0</td>\n",
       "      <td>-18.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>16.0</td>\n",
       "      <td>7.089899</td>\n",
       "      <td>-0.831202</td>\n",
       "      <td>4.440892e-16</td>\n",
       "      <td>63.6075</td>\n",
       "      <td>-62.5350</td>\n",
       "      <td>7.105427e-15</td>\n",
       "      <td>16.0</td>\n",
       "      <td>23.067842</td>\n",
       "      <td>0.079587</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>37.290</td>\n",
       "      <td>-27.555</td>\n",
       "      <td>0.0000</td>\n",
       "      <td>16.0</td>\n",
       "      <td>15.735384</td>\n",
       "      <td>0.404341</td>\n",
       "      <td>1.0</td>\n",
       "      <td>17.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   customer_id year_month  label  cumsum_total_by_cust_id-mean  \\\n",
       "0        12346    2011-10      0                   2953.304457   \n",
       "1        12347    2011-10      1                   2546.704976   \n",
       "2        12348    2011-10      0                   1096.986000   \n",
       "3        12349    2011-10      0                   2014.247369   \n",
       "4        12350    2011-10      0                    281.849118   \n",
       "\n",
       "   cumsum_total_by_cust_id-max  cumsum_total_by_cust_id-min  \\\n",
       "0                  127267.5690                     -85.3710   \n",
       "1                    5613.9435                      29.2050   \n",
       "2                    3332.0100                      21.7800   \n",
       "3                    4367.5335                     -39.8475   \n",
       "4                     551.7600                      41.5800   \n",
       "\n",
       "   cumsum_total_by_cust_id-sum  cumsum_total_by_cust_id-count  \\\n",
       "0                  138805.3095                           47.0   \n",
       "1                  417659.6160                          164.0   \n",
       "2                   55946.2860                           51.0   \n",
       "3                  215524.4685                          107.0   \n",
       "4                    4791.4350                           17.0   \n",
       "\n",
       "   cumsum_total_by_cust_id-std  cumsum_total_by_cust_id-skew  \\\n",
       "0                 18527.811314                      6.855058   \n",
       "1                  1636.394946                      0.120206   \n",
       "2                  1000.019165                      0.812596   \n",
       "3                  1402.866250                      0.089354   \n",
       "4                   161.979220                      0.200043   \n",
       "\n",
       "   cumsum_quantity_by_cust_id-mean  cumsum_quantity_by_cust_id-max  \\\n",
       "0                      1631.063830                         74268.0   \n",
       "1                      1031.786585                          2099.0   \n",
       "2                      1069.450980                          2714.0   \n",
       "3                       467.682243                           988.0   \n",
       "4                       107.941176                           197.0   \n",
       "\n",
       "   cumsum_quantity_by_cust_id-min  cumsum_quantity_by_cust_id-sum  \\\n",
       "0                            10.0                         76660.0   \n",
       "1                             6.0                        169213.0   \n",
       "2                            24.0                         54542.0   \n",
       "3                            -5.0                         50042.0   \n",
       "4                            12.0                          1835.0   \n",
       "\n",
       "   cumsum_quantity_by_cust_id-count  cumsum_quantity_by_cust_id-std  \\\n",
       "0                              47.0                    10825.525671   \n",
       "1                             164.0                      606.713512   \n",
       "2                              51.0                      902.451734   \n",
       "3                             107.0                      273.727436   \n",
       "4                              17.0                       57.957388   \n",
       "\n",
       "   cumsum_quantity_by_cust_id-skew  cumsum_price_by_cust_id-mean  \\\n",
       "0                         6.855636                    400.641415   \n",
       "1                         0.118385                    317.244293   \n",
       "2                         0.493378                     66.215794   \n",
       "3                        -0.077728                    615.014523   \n",
       "4                        -0.168685                     42.133235   \n",
       "\n",
       "   cumsum_price_by_cust_id-max  cumsum_price_by_cust_id-min  \\\n",
       "0                     936.3750                       7.4250   \n",
       "1                     693.4455                       4.8675   \n",
       "2                     318.6150                       0.9075   \n",
       "3                    1484.1585                       6.1875   \n",
       "4                     107.7450                       3.4650   \n",
       "\n",
       "   cumsum_price_by_cust_id-sum  cumsum_price_by_cust_id-count  \\\n",
       "0                   18830.1465                           47.0   \n",
       "1                   52028.0640                          164.0   \n",
       "2                    3377.0055                           51.0   \n",
       "3                   65806.5540                          107.0   \n",
       "4                     716.2650                           17.0   \n",
       "\n",
       "   cumsum_price_by_cust_id-std  cumsum_price_by_cust_id-skew  \\\n",
       "0                   262.090504                      0.440160   \n",
       "1                   207.325091                      0.111363   \n",
       "2                    79.524917                      1.597336   \n",
       "3                   484.322565                      0.447134   \n",
       "4                    40.745863                      0.879343   \n",
       "\n",
       "   cumsum_total_by_prod_id-mean  cumsum_total_by_prod_id-max  \\\n",
       "0                 -18933.867255                  127352.9400   \n",
       "1                  14046.581799                  363807.0975   \n",
       "2                  17500.335471                  146857.4085   \n",
       "3                   8876.528131                   86587.3965   \n",
       "4                  12878.018029                   88027.8630   \n",
       "\n",
       "   cumsum_total_by_prod_id-min  cumsum_total_by_prod_id-sum  \\\n",
       "0                 -145015.2000                -8.898918e+05   \n",
       "1                     156.3375                 2.303639e+06   \n",
       "2                     471.2400                 8.925171e+05   \n",
       "3                      74.2500                 9.497885e+05   \n",
       "4                     292.0500                 2.189263e+05   \n",
       "\n",
       "   cumsum_total_by_prod_id-count  cumsum_total_by_prod_id-std  \\\n",
       "0                           47.0                 64968.445670   \n",
       "1                          164.0                 43323.609176   \n",
       "2                           51.0                 27894.026806   \n",
       "3                          107.0                 12932.316422   \n",
       "4                           17.0                 21329.326160   \n",
       "\n",
       "   cumsum_total_by_prod_id-skew  cumsum_quantity_by_prod_id-mean  \\\n",
       "0                     -1.131166                      2354.212766   \n",
       "1                      6.372546                      4248.969512   \n",
       "2                      3.120231                     11485.372549   \n",
       "3                      3.313619                      2352.869159   \n",
       "4                      3.050607                      3653.352941   \n",
       "\n",
       "   cumsum_quantity_by_prod_id-max  cumsum_quantity_by_prod_id-min  \\\n",
       "0                         74215.0                         -1644.0   \n",
       "1                         45875.0                            56.0   \n",
       "2                         35932.0                           336.0   \n",
       "3                         33089.0                            12.0   \n",
       "4                         16276.0                            60.0   \n",
       "\n",
       "   cumsum_quantity_by_prod_id-sum  cumsum_quantity_by_prod_id-count  \\\n",
       "0                        110648.0                              47.0   \n",
       "1                        696831.0                             164.0   \n",
       "2                        585754.0                              51.0   \n",
       "3                        251757.0                             107.0   \n",
       "4                         62107.0                              17.0   \n",
       "\n",
       "   cumsum_quantity_by_prod_id-std  cumsum_quantity_by_prod_id-skew  \\\n",
       "0                    10767.598390                         6.742285   \n",
       "1                     7357.646380                         3.028850   \n",
       "2                    11056.078952                         0.673806   \n",
       "3                     4244.373963                         4.872195   \n",
       "4                     4338.504177                         1.810547   \n",
       "\n",
       "   cumsum_price_by_prod_id-mean  cumsum_price_by_prod_id-max  \\\n",
       "0                  63650.017691                  325951.2630   \n",
       "1                   1940.714909                   61460.1900   \n",
       "2                   5104.847000                   95332.3635   \n",
       "3                   1555.892636                   30169.8210   \n",
       "4                   3051.218824                   41075.8260   \n",
       "\n",
       "   cumsum_price_by_prod_id-min  cumsum_price_by_prod_id-sum  \\\n",
       "0                       1.6500                 2.991551e+06   \n",
       "1                       8.2500                 3.182772e+05   \n",
       "2                      25.1955                 2.603472e+05   \n",
       "3                      24.2550                 1.664805e+05   \n",
       "4                      77.8800                 5.187072e+04   \n",
       "\n",
       "   cumsum_price_by_prod_id-count  cumsum_price_by_prod_id-std  \\\n",
       "0                           47.0                128903.599171   \n",
       "1                          164.0                  7252.527668   \n",
       "2                           51.0                 16507.639799   \n",
       "3                          107.0                  3418.294904   \n",
       "4                           17.0                  9814.784063   \n",
       "\n",
       "   cumsum_price_by_prod_id-skew  cumsum_total_by_order_id-mean  \\\n",
       "0                      1.619409                      24.297128   \n",
       "1                      6.633689                     470.958796   \n",
       "2                      4.106796                     375.299294   \n",
       "3                      6.236526                     959.884724   \n",
       "4                      4.100398                     281.849118   \n",
       "\n",
       "   cumsum_total_by_order_id-max  cumsum_total_by_order_id-min  \\\n",
       "0                   127352.9400                  -127352.9400   \n",
       "1                     1174.4535                       21.7800   \n",
       "2                     1473.1200                       21.7800   \n",
       "3                     2314.3230                      -39.8475   \n",
       "4                      551.7600                       41.5800   \n",
       "\n",
       "   cumsum_total_by_order_id-sum  cumsum_total_by_order_id-count  \\\n",
       "0                     1141.9650                            47.0   \n",
       "1                    77237.2425                           164.0   \n",
       "2                    19140.2640                            51.0   \n",
       "3                   102707.6655                           107.0   \n",
       "4                     4791.4350                            17.0   \n",
       "\n",
       "   cumsum_total_by_order_id-std  cumsum_total_by_order_id-skew  \\\n",
       "0                  26555.274233                      -0.002867   \n",
       "1                    291.682227                       0.353722   \n",
       "2                    300.618991                       1.639190   \n",
       "3                    703.447214                       0.321784   \n",
       "4                    161.979220                       0.200043   \n",
       "\n",
       "   cumsum_quantity_by_order_id-mean  cumsum_quantity_by_order_id-max  \\\n",
       "0                          3.914894                          74215.0   \n",
       "1                        199.012195                            509.0   \n",
       "2                        388.392157                           1254.0   \n",
       "3                        224.383178                            519.0   \n",
       "4                        107.941176                            197.0   \n",
       "\n",
       "   cumsum_quantity_by_order_id-min  cumsum_quantity_by_order_id-sum  \\\n",
       "0                         -74215.0                            184.0   \n",
       "1                              4.0                          32638.0   \n",
       "2                             24.0                          19808.0   \n",
       "3                             -5.0                          24009.0   \n",
       "4                             12.0                           1835.0   \n",
       "\n",
       "   cumsum_quantity_by_order_id-count  cumsum_quantity_by_order_id-std  \\\n",
       "0                               47.0                     15474.899161   \n",
       "1                              164.0                       134.237770   \n",
       "2                               51.0                       333.951917   \n",
       "3                              107.0                       135.943082   \n",
       "4                               17.0                        57.957388   \n",
       "\n",
       "   cumsum_quantity_by_order_id-skew  cumsum_price_by_order_id-mean  \\\n",
       "0                         -0.000793                      80.737660   \n",
       "1                          0.480564                      57.845881   \n",
       "2                          1.339273                      14.115588   \n",
       "3                          0.244650                     239.373056   \n",
       "4                         -0.168685                      42.133235   \n",
       "\n",
       "   cumsum_price_by_order_id-max  cumsum_price_by_order_id-min  \\\n",
       "0                      412.5990                        1.6500   \n",
       "1                      147.8235                        0.9075   \n",
       "2                       81.8400                        0.4785   \n",
       "3                      793.2540                        2.7225   \n",
       "4                      107.7450                        3.4650   \n",
       "\n",
       "   cumsum_price_by_order_id-sum  cumsum_price_by_order_id-count  \\\n",
       "0                     3794.6700                            47.0   \n",
       "1                     9486.7245                           164.0   \n",
       "2                      719.8950                            51.0   \n",
       "3                    25612.9170                           107.0   \n",
       "4                      716.2650                            17.0   \n",
       "\n",
       "   cumsum_price_by_order_id-std  cumsum_price_by_order_id-skew  order_ts-mean  \\\n",
       "0                     87.485676                       1.486909   1.274484e+09   \n",
       "1                     35.377236                       0.300351   1.297744e+09   \n",
       "2                     18.844181                       2.593528   1.292579e+09   \n",
       "3                    248.595111                       1.326011   1.280047e+09   \n",
       "4                     40.745863                       0.879343   1.296662e+09   \n",
       "\n",
       "   order_ts-max  order_ts-min  order_ts-sum  order_ts-count  order_ts-std  \\\n",
       "0  1.295346e+09  1.260780e+09  5.990076e+10            47.0  8.367102e+06   \n",
       "1  1.312275e+09  1.288535e+09  2.128301e+11           164.0  8.387982e+06   \n",
       "2  1.316956e+09  1.285600e+09  6.592153e+10            51.0  8.085985e+06   \n",
       "3  1.288254e+09  1.259931e+09  1.369650e+11           107.0  8.875333e+06   \n",
       "4  1.296662e+09  1.296662e+09  2.204326e+10            17.0  0.000000e+00   \n",
       "\n",
       "   order_ts-skew  order_ts_diff-mean  order_ts_diff-max  order_ts_diff-min  \\\n",
       "0       0.056381       751438.695652         10197900.0                0.0   \n",
       "1       0.513752       145644.662577          6120780.0                0.0   \n",
       "2       1.617142       627136.800000         14955960.0                0.0   \n",
       "3      -0.411039       267200.377358         14077560.0                0.0   \n",
       "4       0.000000            0.000000                0.0                0.0   \n",
       "\n",
       "   order_ts_diff-sum  order_ts_diff-count  order_ts_diff-std  \\\n",
       "0         34566180.0                 46.0       2.323281e+06   \n",
       "1         23740080.0                163.0       8.397292e+05   \n",
       "2         31356840.0                 50.0       2.475659e+06   \n",
       "3         28323240.0                106.0       1.832081e+06   \n",
       "4                0.0                 16.0       0.000000e+00   \n",
       "\n",
       "   order_ts_diff-skew  quantity_diff-mean  quantity_diff-max  \\\n",
       "0            3.439852        -1613.586957            74216.0   \n",
       "1            5.854925            0.036810              234.0   \n",
       "2            4.710186           -0.460000              138.0   \n",
       "3            7.131391            0.103774               36.0   \n",
       "4            0.000000            0.000000               12.0   \n",
       "\n",
       "   quantity_diff-min  quantity_diff-sum  quantity_diff-count  \\\n",
       "0          -148430.0           -74225.0                 46.0   \n",
       "1             -228.0                6.0                163.0   \n",
       "2             -143.0              -23.0                 50.0   \n",
       "3              -36.0               11.0                106.0   \n",
       "4              -18.0                0.0                 16.0   \n",
       "\n",
       "   quantity_diff-std  quantity_diff-skew  price_diff-mean  price_diff-max  \\\n",
       "0       24684.548109           -4.214331    -1.241087e-01        182.3745   \n",
       "1          27.372576            0.288404     5.263804e-02         20.1300   \n",
       "2          63.018692            0.066540     1.301850e+00         65.5215   \n",
       "3           8.632088           -0.091287    -3.268868e-02        399.3825   \n",
       "4           7.089899           -0.831202     4.440892e-16         63.6075   \n",
       "\n",
       "   price_diff-min  price_diff-sum  price_diff-count  price_diff-std  \\\n",
       "0       -233.1615   -5.709000e+00              46.0       56.879462   \n",
       "1        -16.1700    8.580000e+00             163.0        4.747310   \n",
       "2        -65.5215    6.509250e+01              50.0       24.419082   \n",
       "3       -411.0975   -3.465000e+00             106.0       71.657823   \n",
       "4        -62.5350    7.105427e-15              16.0       23.067842   \n",
       "\n",
       "   price_diff-skew  total_diff-mean  total_diff-max  total_diff-min  \\\n",
       "0        -0.840908     -2770.156304      127360.365     -254705.880   \n",
       "1         0.693660         0.810828         382.635        -387.090   \n",
       "2         0.251831         0.884400         312.840        -327.096   \n",
       "3        -0.213999         0.315212         386.265        -395.670   \n",
       "4         0.079587         0.000000          37.290         -27.555   \n",
       "\n",
       "   total_diff-sum  total_diff-count  total_diff-std  total_diff-skew  \\\n",
       "0    -127427.1900              46.0    42359.030338        -4.214047   \n",
       "1        132.1650             163.0       50.079192         0.000551   \n",
       "2         44.2200              50.0       83.855397        -0.069972   \n",
       "3         33.4125             106.0       75.363508        -0.089535   \n",
       "4          0.0000              16.0       15.735384         0.404341   \n",
       "\n",
       "   order_id-nunique  product_id-nunique  \n",
       "0              17.0                30.0  \n",
       "1               6.0               107.0  \n",
       "2               5.0                25.0  \n",
       "3               4.0                92.0  \n",
       "4               1.0                17.0  "
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "978c7e45",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "categorical feature: []\n",
      "x_tr.shape (5722, 124) , x_te.shape (5914, 124)\n",
      "fold: 1, x_tr.shape: (5149, 121), x_val.shape: (573, 121)\n",
      "Training until validation scores don't improve for 100 rounds\n",
      "Early stopping, best iteration is:\n",
      "[32]\ttraining's auc: 0.94303\tvalid_1's auc: 0.811554\n",
      "Fold 1 | AUC: 0.8115536963684152\n",
      "--------------------------------------------------------------------------------\n",
      "fold: 2, x_tr.shape: (5149, 121), x_val.shape: (573, 121)\n",
      "Training until validation scores don't improve for 100 rounds\n",
      "Early stopping, best iteration is:\n",
      "[20]\ttraining's auc: 0.914591\tvalid_1's auc: 0.796653\n",
      "Fold 2 | AUC: 0.7966534631473736\n",
      "--------------------------------------------------------------------------------\n",
      "fold: 3, x_tr.shape: (5150, 121), x_val.shape: (572, 121)\n",
      "Training until validation scores don't improve for 100 rounds\n",
      "Early stopping, best iteration is:\n",
      "[23]\ttraining's auc: 0.919155\tvalid_1's auc: 0.839524\n",
      "Fold 3 | AUC: 0.8395243784066304\n",
      "--------------------------------------------------------------------------------\n",
      "fold: 4, x_tr.shape: (5150, 121), x_val.shape: (572, 121)\n",
      "Training until validation scores don't improve for 100 rounds\n",
      "Early stopping, best iteration is:\n",
      "[26]\ttraining's auc: 0.92728\tvalid_1's auc: 0.806373\n",
      "Fold 4 | AUC: 0.8063727320241917\n",
      "--------------------------------------------------------------------------------\n",
      "fold: 5, x_tr.shape: (5150, 121), x_val.shape: (572, 121)\n",
      "Training until validation scores don't improve for 100 rounds\n",
      "Early stopping, best iteration is:\n",
      "[55]\ttraining's auc: 0.975263\tvalid_1's auc: 0.805461\n",
      "Fold 5 | AUC: 0.8054612573506224\n",
      "--------------------------------------------------------------------------------\n",
      "fold: 6, x_tr.shape: (5150, 121), x_val.shape: (572, 121)\n",
      "Training until validation scores don't improve for 100 rounds\n",
      "Early stopping, best iteration is:\n",
      "[34]\ttraining's auc: 0.943459\tvalid_1's auc: 0.859721\n",
      "Fold 6 | AUC: 0.8597213719925055\n",
      "--------------------------------------------------------------------------------\n",
      "fold: 7, x_tr.shape: (5150, 121), x_val.shape: (572, 121)\n",
      "Training until validation scores don't improve for 100 rounds\n",
      "Early stopping, best iteration is:\n",
      "[7]\ttraining's auc: 0.867154\tvalid_1's auc: 0.816666\n",
      "Fold 7 | AUC: 0.8166657391433396\n",
      "--------------------------------------------------------------------------------\n",
      "fold: 8, x_tr.shape: (5150, 121), x_val.shape: (572, 121)\n",
      "Training until validation scores don't improve for 100 rounds\n",
      "Early stopping, best iteration is:\n",
      "[8]\ttraining's auc: 0.874198\tvalid_1's auc: 0.807242\n",
      "Fold 8 | AUC: 0.8072421021388688\n",
      "--------------------------------------------------------------------------------\n",
      "fold: 9, x_tr.shape: (5150, 121), x_val.shape: (572, 121)\n",
      "Training until validation scores don't improve for 100 rounds\n",
      "Early stopping, best iteration is:\n",
      "[43]\ttraining's auc: 0.958124\tvalid_1's auc: 0.831024\n",
      "Fold 9 | AUC: 0.8310238002485762\n",
      "--------------------------------------------------------------------------------\n",
      "fold: 10, x_tr.shape: (5150, 121), x_val.shape: (572, 121)\n",
      "Training until validation scores don't improve for 100 rounds\n",
      "Early stopping, best iteration is:\n",
      "[24]\ttraining's auc: 0.922852\tvalid_1's auc: 0.801955\n",
      "Fold 10 | AUC: 0.8019552191737622\n",
      "--------------------------------------------------------------------------------\n",
      "\n",
      "Mean AUC = 0.8176173759994285\n",
      "OOF AUC = 0.8108646752779968\n"
     ]
    }
   ],
   "source": [
    "train, test, y, features = feature_engineering1(df, '2011-12')\n",
    "y_oof, test_preds, fi = make_lgb_oof_prediction(train, y, test, features, model_params=lgb_params)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "id": "18f5c6a2",
   "metadata": {},
   "outputs": [],
   "source": [
    "#fi = plot_feature_importances(fi)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "78f5203d",
   "metadata": {},
   "outputs": [],
   "source": [
    "#Optuna Objective Function 정의\n",
    "def objective(trial, label=label_2011_11):\n",
    "    lgb_params = {\n",
    "        'objective' : 'binary',\n",
    "        'boosting_type' : 'gbdt',\n",
    "        'num_leaves' : trial.suggest_int('num_leaves', 2, 256),\n",
    "        'max_depth' : -1,\n",
    "        'max_bin' : trial.suggest_int('max_bin', 128, 256),\n",
    "        'min_data_in_leaf' : trial.suggest_int('min_data_in_leaf', 10, 40),\n",
    "        'feature_fraction' : trial.suggest_uniform('feature_fraction', 0.4, 1.0),\n",
    "        'bagging_fraction' : trial.suggest_uniform('bagging_fraction', 0.4, 1.0),\n",
    "        'bagging_freq' : trial.suggest_int('bagging_freq', 1, 7),\n",
    "        'n_estimators' : 10000,\n",
    "        'early_stopping_rounds' : 100,\n",
    "        'lambda_l1' : trial.suggest_loguniform('lambda_l1', 1e-8, 10.0),\n",
    "        'lambda_l2' : trial.suggest_loguniform('lambda_l2', 1e-8, 10.0),\n",
    "        'seed' : SEED,\n",
    "        'verbose' : -1,\n",
    "        'n_jobs' : -1,\n",
    "    }\n",
    "    \n",
    "    y_oof, test_preds, fi = make_lgb_oof_prediction(train, y, test, features, model_params=lgb_params)\n",
    "    \n",
    "    val_auc = roc_auc_score(label, y_oof)\n",
    "    \n",
    "    return val_auc\n",
    "\n",
    "def objective_cat(trial, label=label_2011_11):\n",
    "    cat_params = {\n",
    "        'iterations' : trial.suggest_int('iterations', 50, 300),                         \n",
    "        'depth' : trial.suggest_int('depth', 4, 10),                                       \n",
    "        'learning_rate' : trial.suggest_loguniform('learning_rate', 0.01, 0.3),               \n",
    "        'random_strength' :trial.suggest_int('random_strength', 0, 100),                       \n",
    "        'bagging_temperature' :trial.suggest_loguniform('bagging_temperature', 0.01, 100.00),\n",
    "        'learning_rate' :trial.suggest_loguniform('learning_rate', 1e-3, 1e-1),\n",
    "        'od_type': trial.suggest_categorical('od_type', ['IncToDec', 'Iter'])\n",
    "    }\n",
    "    \n",
    "    y_oof, test_preds, fi = make_cat_oof_prediction(train, y, test, features, model_params=cat_params)\n",
    "    \n",
    "    val_auc = roc_auc_score(label, y_oof)\n",
    "    \n",
    "    return val_auc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "823e79f5",
   "metadata": {},
   "outputs": [],
   "source": [
    "#study = optuna.create_study(direction = 'maximize')\n",
    "#study.optimize(objective, n_trials=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "c28ff63e",
   "metadata": {},
   "outputs": [],
   "source": [
    "#study.best_params"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "7513332e",
   "metadata": {},
   "outputs": [],
   "source": [
    "#study.best_value"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "a13e6740",
   "metadata": {},
   "outputs": [],
   "source": [
    "def submission(test_preds, sub_time):\n",
    "    sub = pd.read_csv('input' + '/sample_submission.csv')\n",
    "    sub['probability'] = test_preds\n",
    "    sub.to_csv(os.path.join('../output', sub_time), index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "872b2d58",
   "metadata": {},
   "outputs": [],
   "source": [
    "submission(test_preds, 'sub17.csv')"
   ]
  }
 ],
 "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": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}