{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"import os\nimport random\nimport joblib\nimport gc\nimport itertools\nfrom itertools import combinations\nimport scipy as sp\nimport numpy as np\nimport pandas as pd\nfrom tqdm.notebook import tqdm\n\nimport lightgbm as lgb\nfrom sklearn.preprocessing import LabelEncoder\nfrom hyperopt import STATUS_OK, Trials, fmin, hp, tpe\nfrom sklearn.model_selection import StratifiedKFold, train_test_split\n\n\npd.set_option('display.width', 1000)\npd.set_option('display.max_rows', 500)\npd.set_option('display.max_columns', 500)\nimport warnings; warnings.filterwarnings('ignore')","metadata":{"execution":{"iopub.status.busy":"2022-07-31T07:34:32.903560Z","iopub.execute_input":"2022-07-31T07:34:32.904641Z","iopub.status.idle":"2022-07-31T07:34:35.855591Z","shell.execute_reply.started":"2022-07-31T07:34:32.904513Z","shell.execute_reply":"2022-07-31T07:34:35.853959Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def get_difference(data, num_features):\n    df1 = []\n    customer_ids = []\n    \n    for customer_id, df in tqdm(data.groupby(['customer_ID'])):\n        diff_df1 = df[num_features].diff(1).iloc[[-1]].values.astype(np.float16)#float32をfloat16に変えた\n        df1.append(diff_df1)\n        customer_ids.append(customer_id)\n    \n    df1 = np.concatenate(df1, axis = 0)\n    df1 = pd.DataFrame(df1, columns = [col + '_diff1' for col in df[num_features].columns])\n    df1['customer_ID'] = customer_ids\n    return df1","metadata":{"execution":{"iopub.status.busy":"2022-07-31T07:34:35.859966Z","iopub.execute_input":"2022-07-31T07:34:35.860967Z","iopub.status.idle":"2022-07-31T07:34:35.870871Z","shell.execute_reply.started":"2022-07-31T07:34:35.860908Z","shell.execute_reply":"2022-07-31T07:34:35.869077Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# trainデータの前処理","metadata":{}},{"cell_type":"code","source":"def read_preprocess_traindata(train_, train_labels_):    \n    # 特徴量の指定(customerとS_2以外のすべて)\n    features = train_.drop(['customer_ID', 'target', 'S_2'], axis = 1).columns.to_list()\n    # カテゴリ変数の指定\n    cat_features = [\n        \"B_30\",\n        \"B_38\",\n        \"D_114\",\n        \"D_116\",\n        \"D_117\",\n        \"D_120\",\n        \"D_126\",\n        \"D_63\",\n        \"D_64\",\n        \"D_66\",\n        \"D_68\",\n    ]\n    \n    # 数値変数の個数(特徴量の中からカテゴリ変数で指定したものを除く)\n    num_features = [col for col in features if col not in cat_features]\n    \n    # floatデータのメモリ削減(もともとfloat32だったものをfloat16まで減らす)\n    print('Starting reducing memory. (num feature: cast float32 to float16)')\n    float_cols = list(train_.dtypes[train_.dtypes == 'float32'].index)\n    for col in tqdm(float_cols):\n        train_[col] = train_[col].astype(np.float16)\n    \n    # 特徴量の生成\n    print('Starting training feature engineer...')    \n    train_num_agg = train_.groupby(\"customer_ID\")[num_features].agg(['mean', 'std', 'min', 'max', 'last'])\n    train_num_agg.columns = ['_'.join(x) for x in train_num_agg.columns]\n    train_num_agg.reset_index(inplace = True)\n    train_cat_agg = train_.groupby(\"customer_ID\")[cat_features].agg(['count', 'last', 'nunique'])\n    train_cat_agg.columns = ['_'.join(x) for x in train_cat_agg.columns]\n    train_cat_agg.reset_index(inplace = True)\n    \n    print('Starting reducing memory. (new features:cast 64->8)')\n    # floatデータのメモリ削減(もともとfloat32だったものをfloat16まで減らす)\n    cols = list(train_num_agg.dtypes[train_num_agg.dtypes == 'float64'].index)    \n    for col in tqdm(cols):\n        train_num_agg[col] = train_num_agg[col].astype(np.float16)\n        \n    # intデータのメモリ削減(もともとint32だったものをint8まで減らす)\n    cols = list(train_cat_agg.dtypes[train_cat_agg.dtypes == 'int64'].index)    \n    for col in tqdm(cols):\n        train_cat_agg[col] = train_cat_agg[col].astype(np.int8)\n        \n    print('Starting calculate difference...')\n    train_diff = get_difference(train_, num_features)\n    train_ = train_num_agg.merge(train_cat_agg, how = 'inner', on = 'customer_ID').merge(train_diff, how = 'inner', on = 'customer_ID').merge(train_labels_, how = 'inner', on = 'customer_ID')\n    \n    del train_num_agg, train_cat_agg, train_diff\n    gc.collect()\n    \n    # floatデータのメモリ削減(もともとfloat32だったものをfloat16まで減らす)\n    print('Starting reducing memory. (num feature: cast float16 to float32)')\n    float_cols = list(train_.dtypes[train_.dtypes == 'float16'].index)\n    for col in tqdm(float_cols):\n        train_[col] = train_[col].astype(np.float32)\n    \n    print('Saving file ...')\n    train_.to_parquet('train_fe.parquet')\n    print('Finished Saving file.')","metadata":{"execution":{"iopub.status.busy":"2022-07-31T07:34:35.872783Z","iopub.execute_input":"2022-07-31T07:34:35.873334Z","iopub.status.idle":"2022-07-31T07:34:35.892583Z","shell.execute_reply.started":"2022-07-31T07:34:35.873301Z","shell.execute_reply":"2022-07-31T07:34:35.891471Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# testデータの前処理(trainと同じ)","metadata":{}},{"cell_type":"code","source":"def read_preprocess_testdata(test_):\n    print('Starting test feature engineer...')    \n    test_num_agg = test_.groupby(\"customer_ID\")[num_features].agg(['mean', 'std', 'min', 'max', 'last'])\n    test_num_agg.columns = ['_'.join(x) for x in test_num_agg.columns]\n    test_num_agg.reset_index(inplace = True)\n    test_cat_agg = test_.groupby(\"customer_ID\")[cat_features].agg(['count', 'last', 'nunique'])\n    test_cat_agg.columns = ['_'.join(x) for x in test_cat_agg.columns]\n    test_cat_agg.reset_index(inplace = True)\n    \n    # floatデータのメモリ削減(もともとfloat32だったものをfloat16まで減らす)\n    print('Starting reducing memory. (num feature: cast float32 to float16)')\n    float_cols = list(test_.dtypes[test_.dtypes == 'float32'].index)\n    for col in tqdm(float_cols):\n        test_[col] = test_[col].astype(np.float16)\n    \n    print('Starting reducing memory(cast 64->16) ...')\n    cols = list(test_num_agg.dtypes[test_num_agg.dtypes == 'float64'].index)\n    for col in tqdm(cols):\n        test_num_agg[col] = test_num_agg[col].astype(np.float16)\n        \n    cols = list(test_cat_agg.dtypes[test_cat_agg.dtypes == 'int64'].index)    \n    for col in tqdm(cols):\n        test_cat_agg[col] = test_cat_agg[col].astype(np.int8)\n    \n    print('Starting calculate difference...')\n    test_diff = get_difference(test_, num_features)\n    test_ = test_num_agg.merge(test_cat_agg, how = 'inner', on = 'customer_ID').merge(test_diff, how = 'inner', on = 'customer_ID')\n    \n    del test_num_agg, test_cat_agg, test_diff\n    gc.collect()\n    \n    print('Starting reducing memory. (num feature: cast float16 to float32)')\n    float_cols = list(test_.dtypes[test_.dtypes == 'float16'].index)\n    for col in tqdm(float_cols):\n        test_[col] = test_[col].astype(np.float32)\n    \n    print('Saving file ...')\n    test_.to_parquet('test_fe.parquet')\n    print('Finished Saving file.')","metadata":{"execution":{"iopub.status.busy":"2022-07-31T07:34:35.894803Z","iopub.execute_input":"2022-07-31T07:34:35.895440Z","iopub.status.idle":"2022-07-31T07:34:35.916018Z","shell.execute_reply.started":"2022-07-31T07:34:35.895394Z","shell.execute_reply":"2022-07-31T07:34:35.913571Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# 相関の高いデータはdropする\nDrop_columns_list = ['D_87',\n                     'D_55','D_61','D_74','D_75','D_77',\n                     'D_104','D_107','D_118','D_119',\n                     'D_131','D_132','D_143','D_141',\n                     'S_7',  'S_24',\n                     'B_11', 'B_13', 'B_15', 'B_18',\n                     'B_20', 'B_23', 'B_33', 'B_37',\n                    ]","metadata":{"execution":{"iopub.status.busy":"2022-07-31T07:34:49.384691Z","iopub.execute_input":"2022-07-31T07:34:49.385720Z","iopub.status.idle":"2022-07-31T07:34:49.392005Z","shell.execute_reply.started":"2022-07-31T07:34:49.385678Z","shell.execute_reply":"2022-07-31T07:34:49.390934Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## trainデータの前処理を実行","metadata":{}},{"cell_type":"code","source":"train = pd.read_parquet('../input/amex-parquet/train_data.parquet')\ntrain_labels = pd.read_csv('../input/amex-default-prediction/train_labels.csv')","metadata":{"execution":{"iopub.status.busy":"2022-07-31T07:34:50.629627Z","iopub.execute_input":"2022-07-31T07:34:50.630029Z","iopub.status.idle":"2022-07-31T07:35:29.482071Z","shell.execute_reply.started":"2022-07-31T07:34:50.629996Z","shell.execute_reply":"2022-07-31T07:35:29.478722Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train = train.drop(Drop_columns_list, axis=1)","metadata":{"execution":{"iopub.status.busy":"2022-07-31T07:35:44.886461Z","iopub.execute_input":"2022-07-31T07:35:44.886926Z","iopub.status.idle":"2022-07-31T07:35:49.631538Z","shell.execute_reply.started":"2022-07-31T07:35:44.886890Z","shell.execute_reply":"2022-07-31T07:35:49.630180Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"read_preprocess_traindata(train, train_labels)\ndel train, train_labels","metadata":{"execution":{"iopub.status.busy":"2022-07-31T07:35:51.487793Z","iopub.execute_input":"2022-07-31T07:35:51.488166Z","iopub.status.idle":"2022-07-31T07:40:49.504023Z","shell.execute_reply.started":"2022-07-31T07:35:51.488135Z","shell.execute_reply":"2022-07-31T07:40:49.501390Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]}]}