{"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":"markdown","source":"# Introduction\n\nXGB model with GPU usage by cudf using parameter obtained from Grid Search\n\n\nScored 0.80322 for competition's metrics","metadata":{}},{"cell_type":"markdown","source":"# Load Libraries","metadata":{}},{"cell_type":"code","source":"# LOAD LIBRARIES\nimport pandas as pd, numpy as np # CPU libraries\nimport cupy, cudf # GPU libraries\nimport matplotlib.pyplot as plt, gc, os\nimport seaborn as sns\nimport gc\nimport torch\nfrom numba import cuda\n\nprint('RAPIDS version',cudf.__version__)\npd.set_option('display.max_columns', None)","metadata":{"execution":{"iopub.status.busy":"2023-05-09T14:34:43.141399Z","iopub.execute_input":"2023-05-09T14:34:43.141887Z","iopub.status.idle":"2023-05-09T14:34:51.192312Z","shell.execute_reply.started":"2023-05-09T14:34:43.141842Z","shell.execute_reply":"2023-05-09T14:34:51.191355Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# VERSION NAME FOR SAVED MODEL FILES\nVER = 1\n\n# TRAIN RANDOM SEED\nSEED = 42\n\n# FILL NAN VALUE\nNAN_VALUE = -127 # will fit in int8\n\n# FOLDS PER MODEL\nFOLDS = 5","metadata":{"execution":{"iopub.status.busy":"2023-05-09T14:34:51.194458Z","iopub.execute_input":"2023-05-09T14:34:51.195114Z","iopub.status.idle":"2023-05-09T14:34:51.199442Z","shell.execute_reply.started":"2023-05-09T14:34:51.195080Z","shell.execute_reply":"2023-05-09T14:34:51.198572Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Process and Feature Engineer Train Data\nWe will load @raddar Kaggle dataset from [here][1] with discussion [here][2]. Then we will engineer features suggested by @huseyincot in his notebooks [here][3] and [here][4]. We will use [RAPIDS][5] and the GPU to create new features quickly.\n\n[1]: https://www.kaggle.com/datasets/raddar/amex-data-integer-dtypes-parquet-format\n[2]: https://www.kaggle.com/competitions/amex-default-prediction/discussion/328514\n[3]: https://www.kaggle.com/code/huseyincot/amex-catboost-0-793\n[4]: https://www.kaggle.com/code/huseyincot/amex-agg-data-how-it-created\n[5]: https://rapids.ai/","metadata":{}},{"cell_type":"code","source":"def check_gpu_memory(status = None):\n    \n    context = cuda.current_context()\n    free_memory, total_memory = context.get_memory_info()\n    \n    print(status)\n    print(f'Total memeory{total_memory}')\n    print(f'Free memeory{free_memory}')\n    ","metadata":{"execution":{"iopub.status.busy":"2023-05-09T14:34:51.201159Z","iopub.execute_input":"2023-05-09T14:34:51.201849Z","iopub.status.idle":"2023-05-09T14:34:51.210456Z","shell.execute_reply.started":"2023-05-09T14:34:51.201818Z","shell.execute_reply":"2023-05-09T14:34:51.209594Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def read_file(path = '', usecols = None):\n    # LOAD DATAFRAME\n    if usecols is not None: df = cudf.read_parquet(path, columns=usecols)\n    else: df = cudf.read_parquet(path)\n    # REDUCE DTYPE FOR CUSTOMER AND DATE\n    df['customer_ID'] = df['customer_ID'].str[-16:].str.hex_to_int().astype('int64')\n    df.S_2 = cudf.to_datetime( df.S_2 )\n    # SORT BY CUSTOMER AND DATE (so agg('last') works correctly)\n    df = df.sort_values(['customer_ID','S_2'])\n    df = df.reset_index(drop=True)\n    # FILL NAN\n    df = df.fillna(NAN_VALUE) \n    print('shape of data:', df.shape)\n    \n    return df\n\nprint('Reading train data...')\nTRAIN_PATH = '../input/amex-data-integer-dtypes-parquet-format/train.parquet'\ntrain_raw = read_file(path = TRAIN_PATH)","metadata":{"execution":{"iopub.status.busy":"2023-05-09T14:34:51.213508Z","iopub.execute_input":"2023-05-09T14:34:51.213884Z","iopub.status.idle":"2023-05-09T14:35:14.352305Z","shell.execute_reply.started":"2023-05-09T14:34:51.213857Z","shell.execute_reply":"2023-05-09T14:35:14.351241Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def process_and_feature_engineer(df):\n    \n    gc.collect()\n    torch.cuda.empty_cache()\n    \n    #check_gpu_memory('start')\n    \n    # FEATURE ENGINEERING FROM \n    # https://www.kaggle.com/code/huseyincot/amex-agg-data-how-it-created\n    all_cols = [c for c in list(df.columns) if c not in ['customer_ID','S_2']]\n    cat_features = [\"B_30\",\"B_38\",\"D_114\",\"D_116\",\"D_117\",\"D_120\",\"D_126\",\"D_63\",\"D_64\",\"D_66\",\"D_68\"]\n    num_features = [col for col in all_cols if col not in cat_features]\n    num_diff_features = [ col + '_diff' for col in num_features]\n    \n    test_num_group = df.groupby(\"customer_ID\")\n    \n    #test_num_agg = df.groupby(\"customer_ID\")[num_features].agg(['mean', 'std', 'min', 'max', 'last'])\n    test_num_agg = test_num_group[num_features].agg(['mean', 'std', 'min', 'max', 'last'])\n    test_num_agg.columns = ['_'.join(x) for x in test_num_agg.columns]\n    \n    #Dataframe for diff features\n    df_diff = df[['customer_ID','S_2']].copy()\n    \n    for col in num_features:\n        test_num_agg[f'{col}_max_min_diff'] = test_num_agg[f'{col}_max']-test_num_agg[f'{col}_min']\n        test_num_agg[f'{col}_last_mean_diff'] = (test_num_agg[f'{col}_last']-test_num_agg[f'{col}_mean']).astype('float32')\n        test_num_agg[f'{col}_last_mean_ratio'] = (test_num_agg[f'{col}_last']/test_num_agg[f'{col}_mean']).astype('float32')\n        test_num_agg[f'{col}_min_max_ratio'] = (test_num_agg[f'{col}_min']/test_num_agg[f'{col}_max']).astype('float32')\n        \n        #Extract diff features \n        new_col=col+'_diff'\n        df_diff[new_col]= test_num_group[col].diff().fillna(0)\n\n    test_cat_agg = test_num_group[cat_features].agg(['count', 'last', 'nunique'])\n    #test_cat_agg = df.groupby(\"customer_ID\")[cat_features].agg(['count', 'last', 'nunique'])\n    test_cat_agg.columns = ['_'.join(x) for x in test_cat_agg.columns]\n    \n    df = cudf.concat([test_num_agg, test_cat_agg], axis=1)\n    \n    #check_gpu_memory('after agg')\n    \n    del test_num_agg,test_cat_agg,test_num_group\n    print(df.info())\n    gc.collect()\n    torch.cuda.empty_cache()\n    torch.cuda.synchronize()\n    #check_gpu_memory('after del test_num_agg,test_cat_agg')\n        \n    test_num_diff_agg = df_diff.groupby('customer_ID')[num_diff_features].agg(['max'])\n    #test_num_diff_agg = df.groupby(\"customer_ID\")[num_diff_features].agg(['mean', 'std','max'])\n    test_num_diff_agg.columns = ['_'.join(x) for x in test_num_diff_agg.columns]\n        \n\n    df = cudf.concat([df, test_num_diff_agg], axis=1)\n    \n    #new\n    #df = df.fillna(NAN_VALUE)\n    #df = df.replace(np.Inf,128)\n    #df = df.replace(-np.Inf,-127)\n    \n    del test_num_diff_agg\n    print('shape after engineering', df.shape )\n    \n    return df\n\ntrain = process_and_feature_engineer(train_raw)\ndel train_raw","metadata":{"execution":{"iopub.status.busy":"2023-05-09T14:35:14.354082Z","iopub.execute_input":"2023-05-09T14:35:14.354457Z","iopub.status.idle":"2023-05-09T14:35:27.757337Z","shell.execute_reply.started":"2023-05-09T14:35:14.354424Z","shell.execute_reply":"2023-05-09T14:35:27.756231Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#train.isna().sum().sum()","metadata":{"execution":{"iopub.status.busy":"2023-05-09T14:35:27.759062Z","iopub.execute_input":"2023-05-09T14:35:27.759619Z","iopub.status.idle":"2023-05-09T14:35:27.766981Z","shell.execute_reply.started":"2023-05-09T14:35:27.759568Z","shell.execute_reply":"2023-05-09T14:35:27.765662Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# ADD TARGETS\ntargets = cudf.read_csv('../input/amex-default-prediction/train_labels.csv')\ntargets['customer_ID'] = targets['customer_ID'].str[-16:].str.hex_to_int().astype('int64')\ntargets = targets.set_index('customer_ID')\ntrain = train.merge(targets, left_index=True, right_index=True, how='left')\ntrain.target = train.target.astype('int8')\ndel targets\n\n# NEEDED TO MAKE CV DETERMINISTIC (cudf merge above randomly shuffles rows)\ntrain = train.sort_index().reset_index()\n\n# FEATURES\nFEATURES = train.columns[1:-1]\nprint(f'There are {len(FEATURES)} features!')","metadata":{"execution":{"iopub.status.busy":"2023-05-09T14:35:27.768649Z","iopub.execute_input":"2023-05-09T14:35:27.769083Z","iopub.status.idle":"2023-05-09T14:35:29.520863Z","shell.execute_reply.started":"2023-05-09T14:35:27.769046Z","shell.execute_reply":"2023-05-09T14:35:29.518822Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Train XGB\nWe will train using `DeviceQuantileDMatrix`. This has a very small GPU memory footprint.","metadata":{}},{"cell_type":"code","source":"# LOAD XGB LIBRARY\nfrom sklearn.model_selection import KFold\nimport xgboost as xgb\nprint('XGB Version',xgb.__version__)\n\n# NEEDED WITH DeviceQuantileDMatrix BELOW\nclass IterLoadForDMatrix(xgb.core.DataIter):\n    def __init__(self, df=None, features=None, target=None, batch_size=126*1024):\n        self.features = features\n        self.target = target\n        self.df = df\n        self.it = 0 # set iterator to 0\n        self.batch_size = batch_size\n        self.batches = int( np.ceil( len(df) / self.batch_size ) )\n        super().__init__()\n\n    def reset(self):\n        '''Reset the iterator'''\n        self.it = 0\n\n    def next(self, input_data):\n        '''Yield next batch of data.'''\n        if self.it == self.batches:\n            return 0 # Return 0 when there's no more batch.\n        \n        a = self.it * self.batch_size\n        b = min( (self.it + 1) * self.batch_size, len(self.df) )\n        dt = cudf.DataFrame(self.df.iloc[a:b])\n        input_data(data=dt[self.features], label=dt[self.target]) #, weight=dt['weight'])\n        self.it += 1\n        return 1","metadata":{"execution":{"iopub.status.busy":"2023-05-09T14:35:29.522399Z","iopub.execute_input":"2023-05-09T14:35:29.523274Z","iopub.status.idle":"2023-05-09T14:35:29.851755Z","shell.execute_reply.started":"2023-05-09T14:35:29.523230Z","shell.execute_reply":"2023-05-09T14:35:29.850508Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# https://www.kaggle.com/yunchonggan\n# https://www.kaggle.com/competitions/amex-default-prediction/discussion/328020\ndef amex_metric_mod(y_true, y_pred):\n\n    labels     = np.transpose(np.array([y_true, y_pred]))\n    labels     = labels[labels[:, 1].argsort()[::-1]]\n    weights    = np.where(labels[:,0]==0, 20, 1)\n    cut_vals   = labels[np.cumsum(weights) <= int(0.04 * np.sum(weights))]\n    top_four   = np.sum(cut_vals[:,0]) / np.sum(labels[:,0])\n\n    gini = [0,0]\n    for i in [1,0]:\n        labels         = np.transpose(np.array([y_true, y_pred]))\n        labels         = labels[labels[:, i].argsort()[::-1]]\n        weight         = np.where(labels[:,0]==0, 20, 1)\n        weight_random  = np.cumsum(weight / np.sum(weight))\n        total_pos      = np.sum(labels[:, 0] *  weight)\n        cum_pos_found  = np.cumsum(labels[:, 0] * weight)\n        lorentz        = cum_pos_found / total_pos\n        gini[i]        = np.sum((lorentz - weight_random) * weight)\n    \n    #without name for general  calcaution \n    return 0.5 * (gini[1]/gini[0] + top_four)","metadata":{"execution":{"iopub.status.busy":"2023-05-09T14:35:29.853104Z","iopub.execute_input":"2023-05-09T14:35:29.854206Z","iopub.status.idle":"2023-05-09T14:35:29.863477Z","shell.execute_reply.started":"2023-05-09T14:35:29.854158Z","shell.execute_reply":"2023-05-09T14:35:29.862480Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def amex_metric_mod_xgb(y_pred: np.ndarray, dtrain:xgb.DMatrix):\n\n    y_true     = dtrain.get_label()\n    labels     = np.transpose(np.array([y_true, y_pred]))\n    labels     = labels[labels[:, 1].argsort()[::-1]]\n    weights    = np.where(labels[:,0]==0, 20, 1)\n    cut_vals   = labels[np.cumsum(weights) <= int(0.04 * np.sum(weights))]\n    top_four   = np.sum(cut_vals[:,0]) / np.sum(labels[:,0])\n\n    gini = [0,0]\n    for i in [1,0]:\n        labels         = np.transpose(np.array([y_true, y_pred]))\n        labels         = labels[labels[:, i].argsort()[::-1]]\n        weight         = np.where(labels[:,0]==0, 20, 1)\n        weight_random  = np.cumsum(weight / np.sum(weight))\n        total_pos      = np.sum(labels[:, 0] *  weight)\n        cum_pos_found  = np.cumsum(labels[:, 0] * weight)\n        lorentz        = cum_pos_found / total_pos\n        gini[i]        = np.sum((lorentz - weight_random) * weight)\n\n    #with name to applied in xgboost\n    return 'Amex_eval', 0.5 * (gini[1]/gini[0] + top_four)","metadata":{"execution":{"iopub.status.busy":"2023-05-09T14:35:29.868480Z","iopub.execute_input":"2023-05-09T14:35:29.868840Z","iopub.status.idle":"2023-05-09T14:35:29.881264Z","shell.execute_reply.started":"2023-05-09T14:35:29.868788Z","shell.execute_reply":"2023-05-09T14:35:29.879990Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# XGB MODEL Grid Search PARAMETERS\ngs_xgb_parms = { \n    'max_depth':[3,4,5,6],\n    'gamma': [0.5, 1, 1.5, 2, 5],\n    'learning_rate':[0.05,0.1,0.2,0.3], \n    'subsample':[0.6,0.8,1],\n    'colsample_bytree':[0.6],\n    'objective':['binary:logistic'],\n    'eval_metric':['logloss'],\n    'tree_method':['gpu_hist'],\n    'predictor':['gpu_predictor'],\n    'random_state':[SEED]\n}","metadata":{"execution":{"iopub.status.busy":"2023-05-09T14:35:29.882544Z","iopub.execute_input":"2023-05-09T14:35:29.883038Z","iopub.status.idle":"2023-05-09T14:35:29.896713Z","shell.execute_reply.started":"2023-05-09T14:35:29.883006Z","shell.execute_reply":"2023-05-09T14:35:29.895568Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Grid Search","metadata":{}},{"cell_type":"code","source":"\"\"\"\nfrom sklearn.model_selection import KFold,ParameterGrid\n\nbest_score=0\nfor g in ParameterGrid(gs_xgb_parms):\n    importances = []\n    oof = []\n    #train = train.to_pandas() # free GPU memory\n    TRAIN_SUBSAMPLE = 0.1\n    gc.collect()\n\n    skf = KFold(n_splits=FOLDS, shuffle=True, random_state=SEED)\n    for fold,(train_idx, valid_idx) in enumerate(skf.split(\n                train, train.target )):\n\n        # TRAIN WITH SUBSAMPLE OF TRAIN FOLD DATA\n        if TRAIN_SUBSAMPLE<1.0:\n            np.random.seed(SEED)\n            train_idx = np.random.choice(train_idx, \n                       int(len(train_idx)*TRAIN_SUBSAMPLE), replace=False)\n            np.random.seed(None)\n\n        print('#'*25)\n        print('### Fold',fold+1)\n        print('### Train size',len(train_idx),'Valid size',len(valid_idx))\n        print(f'### Training with {int(TRAIN_SUBSAMPLE*100)}% fold data...')\n        print('#'*25)\n\n        # TRAIN, VALID, TEST FOR FOLD K\n        Xy_train = IterLoadForDMatrix(train.loc[train_idx], FEATURES, 'target')\n        X_valid = train.loc[valid_idx, FEATURES]\n        y_valid = train.loc[valid_idx, 'target']\n\n        dtrain = xgb.DeviceQuantileDMatrix(Xy_train, max_bin=256)\n        dvalid = xgb.DMatrix(data=X_valid, label=y_valid)\n\n        # TRAIN MODEL FOLD K\n        model = xgb.train(g, \n                    dtrain=dtrain,\n                    evals=[(dtrain,'train'),(dvalid,'valid')],\n                    num_boost_round=9999,\n                    early_stopping_rounds=100,\n                    verbose_eval=100) \n        model.save_model(f'XGB_v{VER}_fold{fold}.xgb')\n\n        # GET FEATURE IMPORTANCE FOR FOLD K\n        dd = model.get_score(importance_type='weight')\n        df = pd.DataFrame({'feature':dd.keys(),f'importance_{fold}':dd.values()})\n        importances.append(df)\n\n        # INFER OOF FOLD K\n        oof_preds = model.predict(dvalid)\n        print(type(y_valid.values))\n        print(type(oof_preds))\n        acc = amex_metric_mod(y_valid.values.get(), oof_preds)\n        print('Kaggle Metric =',acc,'\\n')\n\n        # SAVE OOF\n        df = train.loc[valid_idx, ['customer_ID','target'] ].copy()\n        df['oof_pred'] = oof_preds\n        oof.append( df )\n        \n        # save if best\n        if acc > best_score:\n            best_score = acc\n            best_grid = g  \n\n        del dtrain, Xy_train, dd, df\n        del X_valid, y_valid, dvalid, model\n        _ = gc.collect()\n        \n#print('#'*25)\n#oof = cudf.concat(oof,axis=0,ignore_index=True).set_index('customer_ID')\n#acc = amex_metric_mod(oof.target.values.get(), oof.oof_pred.values.get())\n#print('OVERALL CV Kaggle Metric =',acc)        \n        \n        \n        \nprint (\"OOB: %0.5f\" % best_score )\nprint (\"Grid:\", best_grid)\n\"\"\"","metadata":{"execution":{"iopub.status.busy":"2023-05-09T14:35:29.898838Z","iopub.execute_input":"2023-05-09T14:35:29.899568Z","iopub.status.idle":"2023-05-09T14:35:29.913222Z","shell.execute_reply.started":"2023-05-09T14:35:29.899535Z","shell.execute_reply":"2023-05-09T14:35:29.912364Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# best params after grid search\n# XGB MODEL PARAMETERS\nxgb_parms = { \n    'max_depth':3,\n    'gamma': 0.5,\n    'learning_rate':0.05, \n    'subsample':0.8,\n    'eval_metric':'logloss',\n    'colsample_bytree':0.6, \n    'objective':'binary:logistic',\n    'tree_method':'gpu_hist',\n    'predictor':'gpu_predictor',\n    'random_state':SEED\n}","metadata":{"execution":{"iopub.status.busy":"2023-05-09T14:35:29.916118Z","iopub.execute_input":"2023-05-09T14:35:29.916445Z","iopub.status.idle":"2023-05-09T14:35:29.925957Z","shell.execute_reply.started":"2023-05-09T14:35:29.916420Z","shell.execute_reply":"2023-05-09T14:35:29.924948Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"importances = []\noof = []\n#train = train.to_pandas() # free GPU memory\nTRAIN_SUBSAMPLE = 1.0\ngc.collect()\n\nskf = KFold(n_splits=FOLDS, shuffle=True, random_state=SEED)\nfor fold,(train_idx, valid_idx) in enumerate(skf.split(\n            train, train.target )):\n    \n    # TRAIN WITH SUBSAMPLE OF TRAIN FOLD DATA\n    if TRAIN_SUBSAMPLE<1.0:\n        np.random.seed(SEED)\n        train_idx = np.random.choice(train_idx, \n                       int(len(train_idx)*TRAIN_SUBSAMPLE), replace=False)\n        np.random.seed(None)\n    \n    print('#'*25)\n    print('### Fold',fold+1)\n    print('### Train size',len(train_idx),'Valid size',len(valid_idx))\n    print(f'### Training with {int(TRAIN_SUBSAMPLE*100)}% fold data...')\n    print('#'*25)\n    \n    # TRAIN, VALID, TEST FOR FOLD K\n    Xy_train = IterLoadForDMatrix(train.loc[train_idx], FEATURES, 'target')\n    X_valid = train.loc[valid_idx, FEATURES]\n    y_valid = train.loc[valid_idx, 'target']\n    \n    dtrain = xgb.DeviceQuantileDMatrix(Xy_train, max_bin=256)\n    dvalid = xgb.DMatrix(data=X_valid, label=y_valid)\n    \n    # TRAIN MODEL FOLD K\n    model = xgb.train(xgb_parms, \n                dtrain=dtrain,\n                evals=[(dtrain,'train'),(dvalid,'valid')],\n                num_boost_round=9999,\n                early_stopping_rounds=100,\n                verbose_eval=100) \n    model.save_model(f'XGB_v{VER}_fold{fold}.xgb')\n    \n    # GET FEATURE IMPORTANCE FOR FOLD K\n    dd = model.get_score(importance_type='weight')\n    df = pd.DataFrame({'feature':dd.keys(),f'importance_{fold}':dd.values()})\n    importances.append(df)\n            \n    # INFER OOF FOLD K\n    oof_preds = model.predict(dvalid)\n    print(type(y_valid.values))\n    print(type(oof_preds))\n    acc = amex_metric_mod(y_valid.values.get(), oof_preds)\n    print('Kaggle Metric =',acc,'\\n')\n    \n    # SAVE OOF\n    df = train.loc[valid_idx, ['customer_ID','target'] ].copy()\n    df['oof_pred'] = oof_preds\n    oof.append( df )\n    \n    del dtrain, Xy_train, dd, df\n    del X_valid, y_valid, dvalid, model\n    _ = gc.collect()\n    \n#print('#'*25)\n#oof = cudf.concat(oof,axis=0,ignore_index=True).set_index('customer_ID')\n#acc = amex_metric_mod(oof.target.values.get(), oof.oof_pred.values.get())\n#print('OVERALL CV Kaggle Metric =',acc)","metadata":{"scrolled":true,"execution":{"iopub.status.busy":"2023-05-09T14:35:29.929220Z","iopub.execute_input":"2023-05-09T14:35:29.929520Z","iopub.status.idle":"2023-05-09T15:04:55.661332Z","shell.execute_reply.started":"2023-05-09T14:35:29.929487Z","shell.execute_reply":"2023-05-09T15:04:55.660238Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print('#'*25)\noof = cudf.concat(oof,axis=0,ignore_index=True).set_index('customer_ID')\nacc = amex_metric_mod(oof.target.values.get(), oof.oof_pred.values.get())\nprint('OVERALL CV Kaggle Metric =',acc)","metadata":{"execution":{"iopub.status.busy":"2023-05-09T15:04:55.663404Z","iopub.execute_input":"2023-05-09T15:04:55.663760Z","iopub.status.idle":"2023-05-09T15:04:55.956148Z","shell.execute_reply.started":"2023-05-09T15:04:55.663726Z","shell.execute_reply":"2023-05-09T15:04:55.955073Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# CLEAN RAM\ndel train\n_ = gc.collect()","metadata":{"execution":{"iopub.status.busy":"2023-05-09T15:04:55.957726Z","iopub.execute_input":"2023-05-09T15:04:55.958106Z","iopub.status.idle":"2023-05-09T15:04:56.393271Z","shell.execute_reply.started":"2023-05-09T15:04:55.958074Z","shell.execute_reply":"2023-05-09T15:04:56.392239Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Save OOF Preds","metadata":{}},{"cell_type":"code","source":"oof=oof.to_pandas()","metadata":{"execution":{"iopub.status.busy":"2023-05-09T15:04:56.396131Z","iopub.execute_input":"2023-05-09T15:04:56.396546Z","iopub.status.idle":"2023-05-09T15:04:56.409028Z","shell.execute_reply.started":"2023-05-09T15:04:56.396510Z","shell.execute_reply":"2023-05-09T15:04:56.407895Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"oof_xgb = pd.read_parquet(TRAIN_PATH, columns=['customer_ID']).drop_duplicates()\noof_xgb['customer_ID_hash'] = oof_xgb['customer_ID'].apply(lambda x: int(x[-16:],16) ).astype('int64')\noof_xgb = oof_xgb.set_index('customer_ID_hash')\noof_xgb = oof_xgb.merge(oof, left_index=True, right_index=True)\noof_xgb = oof_xgb.sort_index().reset_index(drop=True)\noof_xgb.to_csv(f'oof_xgb_v{VER}.csv',index=False)\noof_xgb.head()","metadata":{"execution":{"iopub.status.busy":"2023-05-09T15:04:56.412281Z","iopub.execute_input":"2023-05-09T15:04:56.413044Z","iopub.status.idle":"2023-05-09T15:05:00.727559Z","shell.execute_reply.started":"2023-05-09T15:04:56.412985Z","shell.execute_reply":"2023-05-09T15:05:00.726379Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# PLOT OOF PREDICTIONS\nplt.hist(oof_xgb.oof_pred.values, bins=100)\nplt.title('OOF Predictions')\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2023-05-09T15:05:00.729249Z","iopub.execute_input":"2023-05-09T15:05:00.729830Z","iopub.status.idle":"2023-05-09T15:05:01.139320Z","shell.execute_reply.started":"2023-05-09T15:05:00.729786Z","shell.execute_reply":"2023-05-09T15:05:01.138209Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# CLEAR VRAM, RAM FOR INFERENCE BELOW\ndel oof_xgb, oof\n_ = gc.collect()","metadata":{"execution":{"iopub.status.busy":"2023-05-09T15:05:01.141276Z","iopub.execute_input":"2023-05-09T15:05:01.141893Z","iopub.status.idle":"2023-05-09T15:05:01.381484Z","shell.execute_reply.started":"2023-05-09T15:05:01.141847Z","shell.execute_reply":"2023-05-09T15:05:01.379926Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Feature Importance","metadata":{}},{"cell_type":"code","source":"import matplotlib.pyplot as plt\n\ndf = importances[0].copy()\nfor k in range(1,FOLDS): df = df.merge(importances[k], on='feature', how='left')\ndf['importance'] = df.iloc[:,1:].mean(axis=1)\ndf = df.sort_values('importance',ascending=False)\ndf.to_csv(f'xgb_feature_importance_v{VER}.csv',index=False)","metadata":{"execution":{"iopub.status.busy":"2023-05-09T15:05:01.384621Z","iopub.execute_input":"2023-05-09T15:05:01.385426Z","iopub.status.idle":"2023-05-09T15:05:01.427106Z","shell.execute_reply.started":"2023-05-09T15:05:01.385387Z","shell.execute_reply":"2023-05-09T15:05:01.426294Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"NUM_FEATURES = 50\nplt.figure(figsize=(10,5*NUM_FEATURES//10))\nplt.barh(np.arange(NUM_FEATURES,0,-1), df.importance.values[:NUM_FEATURES])\nplt.yticks(np.arange(NUM_FEATURES,0,-1), df.feature.values[:NUM_FEATURES])\nplt.title(f'XGB Feature Importance - Top {NUM_FEATURES}')\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2023-05-09T15:05:01.428530Z","iopub.execute_input":"2023-05-09T15:05:01.428851Z","iopub.status.idle":"2023-05-09T15:05:02.141159Z","shell.execute_reply.started":"2023-05-09T15:05:01.428822Z","shell.execute_reply":"2023-05-09T15:05:02.140151Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Process and Feature Engineer Test Data\nWe will load @raddar Kaggle dataset from [here][1] with discussion [here][2]. Then we will engineer features suggested by @huseyincot in his notebooks [here][1] and [here][4]. We will use [RAPIDS][5] and the GPU to create new features quickly.\n\n[1]: https://www.kaggle.com/datasets/raddar/amex-data-integer-dtypes-parquet-format\n[2]: https://www.kaggle.com/competitions/amex-default-prediction/discussion/328514\n[3]: https://www.kaggle.com/code/huseyincot/amex-catboost-0-793\n[4]: https://www.kaggle.com/code/huseyincot/amex-agg-data-how-it-created\n[5]: https://rapids.ai/","metadata":{}},{"cell_type":"code","source":"# CALCULATE SIZE OF EACH SEPARATE TEST PART\ndef get_rows(customers, test, NUM_PARTS = 4, verbose = ''):\n    chunk = len(customers)//NUM_PARTS\n    if verbose != '':\n        print(f'We will process {verbose} data as {NUM_PARTS} separate parts.')\n        print(f'There will be {chunk} customers in each part (except the last part).')\n        print('Below are number of rows in each part:')\n    rows = []\n\n    for k in range(NUM_PARTS):\n        if k==NUM_PARTS-1: cc = customers[k*chunk:]\n        else: cc = customers[k*chunk:(k+1)*chunk]\n        s = test.loc[test.customer_ID.isin(cc)].shape[0]\n        rows.append(s)\n    if verbose != '': print( rows )\n    return rows,chunk\n\n# COMPUTE SIZE OF 4 PARTS FOR TEST DATA\nNUM_PARTS = 4\nTEST_PATH = '../input/amex-data-integer-dtypes-parquet-format/test.parquet'\n\nprint(f'Reading test data...')\ntest = read_file(path = TEST_PATH, usecols = ['customer_ID','S_2'])\ncustomers = test[['customer_ID']].drop_duplicates().sort_index().values.flatten()\nrows,num_cust = get_rows(customers, test[['customer_ID']], NUM_PARTS = NUM_PARTS, verbose = 'test')","metadata":{"execution":{"iopub.status.busy":"2023-05-09T15:05:02.142299Z","iopub.execute_input":"2023-05-09T15:05:02.142583Z","iopub.status.idle":"2023-05-09T15:05:04.758895Z","shell.execute_reply.started":"2023-05-09T15:05:02.142559Z","shell.execute_reply":"2023-05-09T15:05:04.757960Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Infer Test","metadata":{}},{"cell_type":"code","source":"# INFER TEST DATA IN PARTS\nskip_rows = 0\nskip_cust = 0\ntest_preds = []\n\nfor k in range(NUM_PARTS):\n    \n    # READ PART OF TEST DATA\n    print(f'\\nReading test data...')\n    test = read_file(path = TEST_PATH)\n    test = test.iloc[skip_rows:skip_rows+rows[k]]\n    skip_rows += rows[k]\n    print(f'=> Test part {k+1} has shape', test.shape )\n    \n    # PROCESS AND FEATURE ENGINEER PART OF TEST DATA\n    test = process_and_feature_engineer(test)\n    if k==NUM_PARTS-1: test = test.loc[customers[skip_cust:]]\n    else: test = test.loc[customers[skip_cust:skip_cust+num_cust]]\n    skip_cust += num_cust\n    \n    # TEST DATA FOR XGB\n    X_test = test[FEATURES]\n    dtest = xgb.DMatrix(data=X_test)\n    test = test[['P_2_mean']] # reduce memory\n    del X_test\n    gc.collect()\n\n    # INFER XGB MODELS ON TEST DATA\n    model = xgb.Booster()\n    model.load_model(f'XGB_v{VER}_fold0.xgb')\n    preds = model.predict(dtest)\n    for f in range(1,FOLDS):\n        model.load_model(f'XGB_v{VER}_fold{f}.xgb')\n        preds += model.predict(dtest)\n    preds /= FOLDS\n    test_preds.append(preds)\n\n    # CLEAN MEMORY\n    del dtest, model\n    _ = gc.collect()","metadata":{"execution":{"iopub.status.busy":"2023-05-09T15:05:04.760526Z","iopub.execute_input":"2023-05-09T15:05:04.760913Z","iopub.status.idle":"2023-05-09T15:09:34.068252Z","shell.execute_reply.started":"2023-05-09T15:05:04.760878Z","shell.execute_reply":"2023-05-09T15:09:34.067230Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Create Submission CSV","metadata":{}},{"cell_type":"code","source":"# WRITE SUBMISSION FILE\ntest_preds = np.concatenate(test_preds)\ntest = cudf.DataFrame(index=customers,data={'prediction':test_preds})\nsub = cudf.read_csv('../input/amex-default-prediction/sample_submission.csv')[['customer_ID']]\nsub['customer_ID_hash'] = sub['customer_ID'].str[-16:].str.hex_to_int().astype('int64')\nsub = sub.set_index('customer_ID_hash')\nsub = sub.merge(test[['prediction']], left_index=True, right_index=True, how='left')\nsub = sub.reset_index(drop=True)\n\n# DISPLAY PREDICTIONS\nsub.to_csv(f'submission.csv',index=False)\nprint('Submission file shape is', sub.shape )\nsub.head()","metadata":{"execution":{"iopub.status.busy":"2023-05-09T15:09:34.070365Z","iopub.execute_input":"2023-05-09T15:09:34.070710Z","iopub.status.idle":"2023-05-09T15:09:35.361471Z","shell.execute_reply.started":"2023-05-09T15:09:34.070678Z","shell.execute_reply":"2023-05-09T15:09:35.360350Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# PLOT PREDICTIONS\nplt.hist(sub.to_pandas().prediction, bins=100)\nplt.title('Test Predictions')\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2023-05-09T15:09:35.363032Z","iopub.execute_input":"2023-05-09T15:09:35.363882Z","iopub.status.idle":"2023-05-09T15:09:36.237266Z","shell.execute_reply.started":"2023-05-09T15:09:35.363838Z","shell.execute_reply":"2023-05-09T15:09:36.236223Z"},"trusted":true},"execution_count":null,"outputs":[]}]}