{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.10.13","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"gpu","dataSources":[{"sourceId":53666,"databundleVersionId":6589269,"sourceType":"competition"},{"sourceId":6507308,"sourceType":"datasetVersion","datasetId":3706667},{"sourceId":151583600,"sourceType":"kernelVersion"}],"dockerImageVersionId":30698,"isInternetEnabled":false,"language":"python","sourceType":"notebook","isGpuEnabled":true}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"import numpy as np\nimport pandas as pd\nfrom itertools import groupby\nfrom sklearn.model_selection import train_test_split\nfrom pandas.api.types import is_datetime64_ns_dtype\nimport numba\n\nfrom sklearn.ensemble import RandomForestClassifier\nimport xgboost as xgb\nfrom sklearn.ensemble import GradientBoostingClassifier\nfrom sklearn.preprocessing import MinMaxScaler\nimport lightgbm as lgb\nfrom imblearn.under_sampling import RandomUnderSampler\nfrom joblib import Parallel, delayed\nfrom sklearn.model_selection import StratifiedGroupKFold, GroupKFold, TimeSeriesSplit\nfrom sklearn.base import BaseEstimator, RegressorMixin\nfrom sklearn.metrics import average_precision_score\nimport gc\nimport plotly.express as px\nfrom metric import score \n\nimport warnings\nwarnings.filterwarnings(\"ignore\")","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2024-05-21T18:13:00.305514Z","iopub.execute_input":"2024-05-21T18:13:00.306206Z","iopub.status.idle":"2024-05-21T18:13:06.881790Z","shell.execute_reply.started":"2024-05-21T18:13:00.306149Z","shell.execute_reply":"2024-05-21T18:13:06.880956Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"column_names = {\n    'series_id_column_name': 'series_id',\n    'time_column_name': 'step',\n    'event_column_name': 'event',\n    'score_column_name': 'score',\n}\n\ntolerances = {\n    'onset': [12, 36, 60, 90, 120, 150, 180, 240, 300, 360], \n    'wakeup': [12, 36, 60, 90, 120, 150, 180, 240, 300, 360]\n}","metadata":{"execution":{"iopub.status.busy":"2024-05-21T18:13:06.883326Z","iopub.execute_input":"2024-05-21T18:13:06.883931Z","iopub.status.idle":"2024-05-21T18:13:06.889083Z","shell.execute_reply.started":"2024-05-21T18:13:06.883904Z","shell.execute_reply":"2024-05-21T18:13:06.888256Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def reduce_mem_usage(df):\n    \n    \"\"\" \n    Iterate through all numeric columns of a dataframe and modify the data type\n    to reduce memory usage.   \n    \"\"\"\n    start_mem = df.memory_usage().sum() / 1024**2\n    print(f'Memory usage of dataframe is {start_mem:.2f} MB')\n    \n    for col in df.columns:\n        col_type = df[col].dtype\n\n        if col_type != object and not is_datetime64_ns_dtype(df[col]) and not 'category':\n            c_min = df[col].min()\n            c_max = df[col].max()\n            if str(col_type)[:3] == 'int':\n                if c_min >= 0:\n                    if c_min > np.iinfo(np.uint8).min and c_max < np.iinfo(np.uint8).max:\n                        df[col] = df[col].astype(np.uint8)\n                    elif c_min > np.iinfo(np.uint16).min and c_max < np.iinfo(np.uint16).max:\n                        df[col] = df[col].astype(np.uint16)\n                    elif c_min > np.iinfo(np.uint32).min and c_max < np.iinfo(np.uint32).max:\n                        df[col] = df[col].astype(np.uint32)\n                    elif c_min > np.iinfo(np.uint64).min and c_max < np.iinfo(np.uint64).max:\n                        df[col] = df[col].astype(np.uint64)  \n                else:\n                    if c_min > np.iinfo(np.int8).min and c_max < np.iinfo(np.int8).max:\n                        df[col] = df[col].astype(np.int8)\n                    elif c_min > np.iinfo(np.int16).min and c_max < np.iinfo(np.int16).max:\n                        df[col] = df[col].astype(np.int16)\n                    elif c_min > np.iinfo(np.int32).min and c_max < np.iinfo(np.int32).max:\n                        df[col] = df[col].astype(np.int32)\n                    elif c_min > np.iinfo(np.int64).min and c_max < np.iinfo(np.int64).max:\n                        df[col] = df[col].astype(np.int64)   \n            else:\n                if c_min > np.finfo(np.float16).min and c_max < np.finfo(np.float16).max:\n                    df[col] = df[col].astype(np.float16)\n                elif c_min > np.finfo(np.float32).min and c_max < np.finfo(np.float32).max:\n                    df[col] = df[col].astype(np.float32)\n                else:\n                    df[col] = df[col].astype(np.float64)\n        \n    print(f'Memory usage after optimization is: {end_mem:.2f} MB')\n    decrease = 100 * (start_mem - end_mem) / start_mem\n    print(f'Decreased by {decrease:.2f}%')\n    \n    gc.collect()\n    \n    return df","metadata":{"execution":{"iopub.status.busy":"2024-05-21T18:13:06.890186Z","iopub.execute_input":"2024-05-21T18:13:06.890443Z","iopub.status.idle":"2024-05-21T18:13:06.907522Z","shell.execute_reply.started":"2024-05-21T18:13:06.890420Z","shell.execute_reply":"2024-05-21T18:13:06.906680Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# **Feature Engineering**","metadata":{}},{"cell_type":"code","source":"def get_macd(s, hfast=4, hslow=48, macd=15, center=False):\n    \n    sma_fast = s.rolling(hfast*12*60, min_periods=1, center=center).agg('mean')\n    sma_slow = s.rolling(hslow*12*60, min_periods=1, center=center).agg('mean')\n    macd = (sma_fast - sma_slow).rolling(12*macd, min_periods=1, center=center).mean().astype(np.float32)\n    \n    return macd","metadata":{"execution":{"iopub.status.busy":"2024-05-21T18:13:06.910067Z","iopub.execute_input":"2024-05-21T18:13:06.910660Z","iopub.status.idle":"2024-05-21T18:13:06.920719Z","shell.execute_reply.started":"2024-05-21T18:13:06.910630Z","shell.execute_reply":"2024-05-21T18:13:06.919889Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def cls_zone(df, col='anglezdiffabs'):\n    \n    s = df[f'{col}']\n    df[f'{col}_macd'] = get_macd(s)\n\n    s = df[f'{col}'].sort_index(ascending=False)\n    df[f'{col}_macd_rev'] = get_macd(s).sort_index()\n    \n    df[f'{col}_macd_spred'] = df[f'{col}_macd']-df[f'{col}_macd_rev']\n    df[f'{col}_macd_spread_diff'] =df[f'{col}_macd_spred'].diff(60)\n    \n    df.loc[df[f'{col}_macd_spred']<3, 'wakeup_zone'] = 1\n    df['wakeup_zone']=df['wakeup_zone'].fillna(0)\n    df.loc[df[f'{col}_macd_spred']>-3, 'onset_zone'] = 1\n    df['onset_zone']=df['onset_zone'].fillna(0)\n    \n    return df","metadata":{"execution":{"iopub.status.busy":"2024-05-21T18:13:06.921803Z","iopub.execute_input":"2024-05-21T18:13:06.922543Z","iopub.status.idle":"2024-05-21T18:13:06.931542Z","shell.execute_reply.started":"2024-05-21T18:13:06.922512Z","shell.execute_reply":"2024-05-21T18:13:06.930577Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def feat_eng(df):\n    \n    df['series_id'] = df['series_id'].astype('category')\n    df['timestamp'] = pd.to_datetime(df['timestamp'], format='%Y-%m-%dT%H:%M:%S%z',utc=False).apply(lambda t: t.tz_localize(None))\n    df['hour'] = df[\"timestamp\"].dt.hour\n    df['month'] = df[\"timestamp\"].dt.month\n    \n    df.sort_values(['timestamp'], inplace=True)\n    df.set_index('timestamp', inplace=True)\n    \n    # limit outiliers in enmo\n    df['enmo'] = df['enmo'].clip(upper=4)\n    \n    df['lids'] = np.maximum(0., df['enmo'] - 0.02)\n    df['lids'] = df['lids'].rolling(f'{120*5}s', center=True, min_periods=1).agg('sum')\n    df['lids'] = 100 / (df['lids'] + 1)\n    df['lids'] = df['lids'].rolling(f'{360*5}s', center=True, min_periods=1).agg('mean').astype(np.float32)\n    \n    df['lids_sma_fast'] = df['lids'].rolling('2h', center=True, min_periods=1).agg('mean').astype(np.float32)\n    df['lids_sma_macd'] = (df['lids'].rolling('18h', center=True, min_periods=1).agg('mean') - df['lids_sma_fast']\n                                 ).rolling(f'1h', center=True, min_periods=1).agg('mean').astype(np.float32)\n    \n    df[\"enmo\"] = (df[\"enmo\"]*1000).astype(np.int16)\n    df[\"anglez\"] = df[\"anglez\"].astype(np.int16)\n    df[\"anglezdiffabs\"] = df[\"anglez\"].diff().abs().astype(np.float32)\n    \n    df = cls_zone(df, col='anglezdiffabs')\n\n    for col in ['enmo', 'anglez', 'anglezdiffabs']:\n        \n        df[f'{col}_sma_fast'] = df[f'{col}'].rolling(f'4h', center=True, min_periods=1).agg('mean').astype(np.float32)\n        df[f'{col}_sf_macd'] = (df[f'{col}_sma_fast'] - df[f'{col}'].rolling(f'48h', center=True, min_periods=1).agg('mean')\n                                   ).rolling(f'1h', center=True, min_periods=1).agg('mean').astype(np.float32)\n        \n        # periods in seconds        \n        periods = [60, 300, 720, 1440, 3600] \n        \n        for n in periods:\n            \n            rol_args = {'window':f'{n+5}s', 'min_periods':10, 'center':True}\n            \n            for agg in ['median', 'mean', 'max', 'min', 'std']:\n                df[f'{col}_{agg}_{n}'] = df[col].rolling(**rol_args).agg(agg).astype(np.float32).values\n                gc.collect()\n            \n            if n == max(periods):\n                df[f'{col}_mad_{n}'] = (df[col] - df[f'{col}_median_{n}']).abs().rolling(**rol_args).median().astype(np.float32)\n            \n            df[f'{col}_amplit_{n}'] = df[f'{col}_max_{n}']-df[f'{col}_min_{n}']\n            df[f'{col}_amplit_{n}_min'] = df[f'{col}_amplit_{n}'].rolling(**rol_args).min().astype(np.float32).values\n            \n            df[f'{col}_diff_{n}_max'] = df[f'{col}_max_{n}'].diff().abs().rolling(**rol_args).max().astype(np.float32)\n            df[f'{col}_diff_{n}_mean'] = df[f'{col}_max_{n}'].diff().abs().rolling(**rol_args).mean().astype(np.float32)\n\n    \n            gc.collect()\n        \n    #df.drop(columns=[], inplace=True)\n    \n    rol_args = {'window':f'8h', 'min_periods':15, 'center':True}\n    df['not_wear'] = df[f'anglez_std_720'].rolling(**rol_args).max().astype(np.float32)\n    \n    \n    df.reset_index(inplace=True)\n    df.dropna(inplace=True)\n    \n    df = df[(df['wakeup_zone']==1)|(df['onset_zone']==1)]\n    \n\n    return df","metadata":{"execution":{"iopub.status.busy":"2024-05-21T18:13:06.933020Z","iopub.execute_input":"2024-05-21T18:13:06.933592Z","iopub.status.idle":"2024-05-21T18:13:06.953636Z","shell.execute_reply.started":"2024-05-21T18:13:06.933539Z","shell.execute_reply":"2024-05-21T18:13:06.952892Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"file = '/kaggle/input/zzzs-lightweight-training-dataset-target/Zzzs_train.parquet'\n\ndef feat_eng_by_id(idx):\n    \n    from warnings import simplefilter \n    simplefilter(action=\"ignore\", category=pd.errors.PerformanceWarning)\n    \n    df  = pd.read_parquet(file, filters=[('series_id','=',idx)])\n    df['awake'] = df['awake'].astype(np.int8)\n    df = feat_eng(df)\n    \n    return df","metadata":{"execution":{"iopub.status.busy":"2024-05-21T18:13:06.954538Z","iopub.execute_input":"2024-05-21T18:13:06.954796Z","iopub.status.idle":"2024-05-21T18:13:06.966067Z","shell.execute_reply.started":"2024-05-21T18:13:06.954775Z","shell.execute_reply":"2024-05-21T18:13:06.965218Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"DEV = False\n\nseries_id  = pd.read_parquet(file, columns=['series_id'])\nseries_id = series_id.series_id.unique()\n\nprint('nunique id', len(series_id))\n\nif DEV:\n    series_id = series_id[::10]\n    print('DEV nunique ids:', len(series_id))","metadata":{"execution":{"iopub.status.busy":"2024-05-21T18:13:06.967216Z","iopub.execute_input":"2024-05-21T18:13:06.967563Z","iopub.status.idle":"2024-05-21T18:13:09.077858Z","shell.execute_reply.started":"2024-05-21T18:13:06.967533Z","shell.execute_reply":"2024-05-21T18:13:09.076911Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"weird_series = ['31011ade7c0a', 'a596ad0b82aa']\n\nseries_id = [s for s in series_id if s not in weird_series]","metadata":{"execution":{"iopub.status.busy":"2024-05-21T18:13:09.079073Z","iopub.execute_input":"2024-05-21T18:13:09.079416Z","iopub.status.idle":"2024-05-21T18:13:09.083885Z","shell.execute_reply.started":"2024-05-21T18:13:09.079390Z","shell.execute_reply":"2024-05-21T18:13:09.082939Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"%%time\n\ntrain = Parallel(n_jobs=-1)(delayed(feat_eng_by_id)(i) for i in series_id)\ntrain = pd.concat(train, ignore_index=True)","metadata":{"execution":{"iopub.status.busy":"2024-05-21T18:13:09.087628Z","iopub.execute_input":"2024-05-21T18:13:09.087982Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train.shape","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train = train.loc[train['not_wear']>22]\ntrain.shape","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"drop_cols = ['series_id', 'step', 'timestamp']\ngroup = train['series_id']\nX, y = train.drop(columns=drop_cols+['awake']), train['awake']\n# reduce train dataset by half\nstep = 60\nX, y = X.iloc[::step], y[::step]\ngroup = group[::step]\nstep","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"y.value_counts()","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"del train\ngc.collect()","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"X.head()","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# **Ensemble Model**","metadata":{}},{"cell_type":"code","source":"lgb_params = {    \n    'boosting_type':'gbdt',\n    #'data_sample_strategy': 'goss' ,\n    'num_leaves':64,\n    'max_depth':20,\n    'device':'gpu',\n    'learning_rate':0.03,\n    'n_estimators':500,\n    'subsample_for_bin':200000,\n    'min_child_weight':0.001,\n    'min_child_samples':20,\n    'subsample':0.8,\n    'metric':'average_precision',\n    #'colsample_bytree':0.7,\n    'reg_alpha':0.025,\n    'reg_lambda':0.025,\n    'random_state':21520398\n             }","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#tscv = TimeSeriesSplit(n_splits=5)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"%%time\n#fitted_models = []\n#cv_scores = []\n\nclf_lgm = lgb.LGBMClassifier(**lgb_params)\nclf_lgm.fit(\n    X, y,\n    #eval_set=[(X_valid, y_valid)],\n    #callbacks = [lgb.log_evaluation(200), lgb.early_stopping(400)]\n)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# **Features importance** ","metadata":{}},{"cell_type":"code","source":"feats = []\n\nfor m in [clf_lgm]:\n    feat_imp = m.feature_importances_\n    \n    feat_imp = MinMaxScaler().fit_transform(feat_imp.reshape(-1, 1))\n    feat_imp = pd.Series(pd.Series(feat_imp.reshape(1, -1)[0], index=X.columns).sort_values(), index=X.columns).sort_values()\n    feats.append(feat_imp)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"feat_imp = pd.Series(pd.concat(feats, axis=1).mean(axis=1), index=feats[0].index).sort_values()\nfig = px.bar(x=feat_imp, y=feat_imp.index, orientation='h')\nfig.show()","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print('Columns with poor contribution', feat_imp[feat_imp<0.005].index, sep='\\n')","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"del X,y\ngc.collect()","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"feat_imp.sort_values().head(10)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"feat_imp.sort_values().tail(10)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def get_events(idx, classifier, file='test_series.parquet') :\n    \n    test  = pd.read_parquet(f'/kaggle/input/child-mind-institute-detect-sleep-states/{file}',\n                    filters=[('series_id','=',idx)])\n    \n    test = feat_eng(test)\n    \n    X_test = test.drop(columns=drop_cols)\n    \n    if test.shape[0]>0:\n\n        preds, probs = classifier.predict(X_test), classifier.predict_proba(X_test)[:, 1]\n\n        test['prediction'] = preds\n        test['prediction'] = test['prediction'].rolling(60+1, center=True).median()\n        test['probability'] = probs\n\n        test = test[test['prediction']!=2]\n\n        test.loc[test['prediction']==0, 'probability'] = 1-test.loc[test['prediction']==0, 'probability']\n        test['score'] = test['probability'].rolling(60*12*5, center=True, min_periods=10).mean().bfill().ffill()\n        test['pred_diff'] = test['prediction'].diff()\n        \n        test['event'] = test['pred_diff'].replace({1:'wakeup', -1:'onset', 0:np.nan})\n        test.dropna(subset='event', inplace=True)\n        test['step_diff'] = test['step'].diff()\n        test = test.loc[test['not_wear']>20]\n        test.sort_values('step_diff', inplace=True)\n        \n\n        test['day'] = ((test['step']//17280)+1).astype(np.int32)\n        \n        \n        mask_wk= (test['event']=='wakeup') #& (test['wakeup_zone']==1)\n        test_wakeup = test[mask_wk].groupby('day', as_index=False).agg('last')\n        \n        mask_os= (test['event']=='onset') #& (test['onset_zone']==1)\n        test_onset = test[mask_os].groupby('day', as_index=False).agg('last')\n\n        \n        df_events = pd.concat([test_wakeup, test_onset], ignore_index=True).sort_values('step')\n        \n        \n        if df_events.shape[0]>0:\n    \n            if df_events.iloc[0]['event'] == 'wakeup':\n                df_events = df_events.iloc[1:]\n\n    return df_events","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"%%time\n\ncols_sub = ['series_id','step','event','score']\n\nseries_id  = pd.read_parquet('/kaggle/input/child-mind-institute-detect-sleep-states/test_series.parquet', columns=['series_id'])\nseries_id = series_id.series_id.unique()\n\ntests = Parallel(n_jobs=-1)(delayed(get_events)(i, clf_lgm) for i in series_id)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"submission = pd.concat(tests, ignore_index=True)\nsubmission = submission[cols_sub].reset_index(names='row_id')\nsubmission.to_csv('submission.csv', index=False)\nsubmission","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"%%time\n\nprint('validation')\n\n#series_id  = pd.read_parquet('/kaggle/input/child-mind-institute-detect-sleep-states/train_series.parquet', columns=['series_id'])\n#series_id = series_id.series_id.unique()\n\npath = '/kaggle/input/child-mind-institute-detect-sleep-states/'\n\ntrain_events = pd.read_csv(path + 'train_events.csv')\ntrain_events['timestamp'] = pd.to_datetime(train_events['timestamp']).apply(lambda t: t.tz_localize(None))\n\nval_solution = train_events[~train_events['series_id'].isin(series_id)]\nseries_id = train_events.series_id.unique()\n\nvals = Parallel(n_jobs=-1)(delayed(get_events)(i, clf_lgm, file='train_series.parquet') for i in series_id)\n\nval=pd.concat(vals, ignore_index=True).reset_index(names='row_id')\n\nprint(f\"LGBM score: {score(val_solution, val, tolerances, **column_names)}\")\n\nval['score'] = 1\nprint(f\"LGBM score: {score(val_solution, val, tolerances, **column_names)}\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]}]}