{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.10.14","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"none","dataSources":[{"sourceId":81933,"databundleVersionId":9643020,"sourceType":"competition"},{"sourceId":9695159,"sourceType":"datasetVersion","datasetId":5899869},{"sourceId":203049136,"sourceType":"kernelVersion"}],"dockerImageVersionId":30786,"isInternetEnabled":false,"language":"python","sourceType":"notebook","isGpuEnabled":false}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"**The goal** of this competition is to predict children's problematic internet use. The main target is Severity Impairment Index ('sii') that consists of 4 classes (0=None to 3=Severe) and is derived from the sum (PCIAT-PCIAT_Total) of 20 Parent-Child Internet addiction tests (PCIAT-PCIAT_01 to 20). Separation of 'PCIAT-PCIAT_Total' into 'sii' classes is made by the thresholds: 0-30=None, 31-49=Mild, 50-79=Moderate, 80-100=Severe. Each individual 'PCIAT-PCIAT' test consists of 0 to 5 points value that added up together result in 'PCIAT-PCIAT_Total'.\n\nAvailable data comprises one main dataset with different information about participants and measurements from a variety of instruments and questionnaires. In addition to this - a part of participants worn a biotracker that registered a time-series with actigraphy measurements (different acceleration and positional data). \n\n**Data process**\n\nFor efficiency, we will use Polars to read and aggregate all the actigraphy time-series and will convert a long time-series dataframe into a wide format that will permit to join it to each individual participant's 'id' from the main dataset.\n\nBy using an autoencoder we will reduce the high dimensionality of the wide format (more than 260 columns) of aggregated series files to a lower dimensionality and will break the colliniarity between aggregated features.\n\n**Model**\n\nEven if it's a classification problem on the surface, we will use regression models because the target has an ordinal structure (increasing from None to Severe).\n\nFor prediction we will build a deep (multi-level) ensemble of gradient boosting models:\n\n***The first group of models:*** 3 single regressors (tuned LGBM, XGB and CatBoost) will be trained on the joined dataframes (main + series aggregations) and will predict the 'sii' class. Class separation will be based on optimized thresholds (using Scipy minimize).\n\n***The second group of models (60 in total):*** 3 multi output regressors with tuned LGBM, XGB and CatBoost as base estimators will be also trained on the full train set and will predict each of 20 'PCIAT-PCIAT' tests. All these predictions will be added together to obtain a 'PCIAT-PCIAT_Total' value, that at it's turn will be divided into 4 'sii' classes based on optimized thresholds.\n\n***The third group of models:*** 3 single regressors (tuned LGBM, XGB and CatBoost) will be trained on meta features represented by the raw predictions from the second group of models (60 meta features from 60 models) and will predict the 'sii' class directly. Class separation will be also based on optimized custom thresholds.\n\n***As a last layer:*** predictions from all 3 groups will be averaged and rounded to a final 'sii' class result.\n\n![napkin-selection_dark_back.jpg](attachment:73036664-7a33-4196-9324-ac6f410f51c3.jpg)","metadata":{},"attachments":{"73036664-7a33-4196-9324-ac6f410f51c3.jpg":{"image/jpeg":"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"}}},{"cell_type":"code","source":"import pandas as pd\nimport numpy as np\nimport polars as pl\nimport lightgbm as lgb\nimport xgboost as xgb\nimport catboost as cb\nimport matplotlib.pyplot as plt\nimport seaborn as sns\nimport os, glob, optuna, torch, warnings\nwarnings.simplefilter(action='ignore', category=FutureWarning)\npd.set_option('display.max_colwidth', 100)\n\nfrom tqdm import tqdm\nfrom sklearn.impute import SimpleImputer\nfrom sklearn.preprocessing import OneHotEncoder, QuantileTransformer                          \nfrom sklearn.pipeline import make_pipeline\nfrom sklearn.compose import make_column_transformer\nfrom sklearn.utils import shuffle\nfrom sklearn.multioutput import MultiOutputRegressor\nfrom sklearn.model_selection import (\n    cross_val_predict, StratifiedKFold, train_test_split)\nfrom sklearn.metrics import (\n    make_scorer, cohen_kappa_score, confusion_matrix)\nfrom sklearn.cluster import KMeans\nfrom sklearn.mixture import GaussianMixture\nfrom scipy.optimize import minimize\nfrom keras.models import Model\nfrom keras.layers import Input, Dense\nfrom keras.optimizers import Adam","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2024-12-03T11:48:22.383012Z","iopub.execute_input":"2024-12-03T11:48:22.384197Z","iopub.status.idle":"2024-12-03T11:48:44.444403Z","shell.execute_reply.started":"2024-12-03T11:48:22.38414Z","shell.execute_reply":"2024-12-03T11:48:44.443281Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"### Aggregation and pre-process functions\n\ndef create_series_df_base(series_path):\n    '''\n    Using Polars read and aggregate all the series files at once. \n    The 'id' from each file name will be extracted into a separate \n    column because Polars recognizes the hive data structure of the folder  \n    '''\n    sign_cols = ['X', 'Y', 'Z', 'anglez']\n    pl_df = (\n        pl.scan_parquet(series_path)\n        .drop('step')\n        # Exclude values when the tracker wasn't worn\n        .filter(pl.col('non-wear_flag') != 1)\n        .select(pl.exclude(['non-wear_flag', 'battery_voltage']))\n        .with_columns(\n            # Convert to time format\n            [pl.col('time_of_day').cast(pl.Time).alias('time'),\n             # Separate day into 4 parts\n             pl.col('time_of_day').cut(\n                 [7*3600*1e9, 12*3600*1e9, 17*3600*1e9, 22*3600*1e9], \n                 labels=['night', 'morning', 'af_noon', 'evening', 'night'])\n             .cast(pl.Categorical).alias('day_part'),\n             # Light to enmo ratios  \n             (pl.col('light') / (pl.col('enmo') + 1e-5)).cast(pl.Float32)\n             .alias('light_enmo_ratio'),\n             # Light to anglez ratios \n             (pl.col('light') / (pl.col('anglez') + 1e-5)).cast(pl.Float32)\n             .alias('light_anglez_ratio'),\n             # Movement with a specific arm orientation\n             (pl.col('enmo') * pl.col('anglez')).cast(pl.Float32)\n             .alias('enmo_anglez_move'),\n            ]\n            # Calculate the cosine of the angles with respect to each axis\n            + [(pl.col(c) / (pl.col('enmo') + 1e-5)).cast(pl.Float32)\n               .alias(f'cos_angle_{c}') for c in ['X', 'Y', 'Z']]\n            # Marker of values' positive or negative sign\n            + [pl.col(c).sign().cast(pl.Int8).alias(f'{c}_sign') for c in sign_cols]\n        ))\n    # Average the cosine of the angles\n    pl_df = (pl_df.with_columns(\n        ((pl.col('cos_angle_X') + pl.col('cos_angle_Y') + pl.col('cos_angle_Z')) / 3)\n        .cast(pl.Float32).alias('cos_avg'))\n    )\n    return pl_df\n\ndef day_4parts_agg_p1(pl_df):\n    '''\n    Part 1\n    For every id calculate the average of aggregations (by part of the day)\n    '''\n    all_cols = ['X', 'Y', 'Z', 'enmo', 'anglez', 'light']\n    min_cols = ['X', 'Y', 'Z', 'anglez']\n    abs_mean_cols = min_cols\n    range_cols = min_cols\n    sign_cols = ['X_sign', 'Y_sign', 'Z_sign', 'anglez_sign']\n    \n    pl_df = (\n        pl_df.group_by(['id', 'day_part'])\n        .agg(\n            [pl.col(c).cast(pl.Float32).mean().alias(f'{c}_mean') for c in all_cols]\n            + [pl.col(c).cast(pl.Float32).abs().mean().alias(f'{c}_abs_mean') \n               for c in abs_mean_cols]\n            + [pl.col(c).cast(pl.Float32).median().alias(f'{c}_median') \n               for c in all_cols]\n            + [pl.col(c).cast(pl.Float32).min().alias(f'{c}_min') for c in min_cols]\n            + [pl.col(c).cast(pl.Float32).max().alias(f'{c}_max') for c in all_cols] \n            + [pl.col(c).cast(pl.Float32).std().alias(f'{c}_std') for c in all_cols]\n            + [pl.col(c).cast(pl.Float32).quantile(0.25).alias(f'{c}_q25') \n               for c in all_cols]\n            + [pl.col(c).cast(pl.Float32).quantile(0.75).alias(f'{c}_q75') \n               for c in all_cols]\n            # Range of values (max - min)\n            + [(pl.col(c).max() - pl.col(c).min()).cast(pl.Float32).alias(f'{c}_range') \n               for c in range_cols]\n            # Number of sign changes (fluctuations)\n            + [(pl.col(c).shift(1) != pl.col(c)).cast(pl.Int32).sum()\n               .alias(f'{c}_change_sum') for c in sign_cols]\n            # How many times enmo exceeds the overall average (count of spikes)\n            + [(pl.col('enmo') > 0.05).cast(pl.Int32).sum().alias('high_enmo_sum')]\n        )\n        .collect()\n        .pivot('day_part', index='id', aggregate_function='mean', sort_columns=True)\n    )\n    return pl_df\n\ndef day_4parts_agg_p2(pl_df):\n    '''\n    Part 2\n    For every id calculate the average of aggregations (by part of the day) \n    '''\n    all_cols = ['light_enmo_ratio', 'light_anglez_ratio', 'enmo_anglez_move', \n                'cos_avg']\n    pl_df = (\n        pl_df.group_by(['id', 'day_part'])\n        .agg(\n            [pl.col(c).cast(pl.Float32).quantile(0.25).alias(f'{c}_q25') \n               for c in all_cols]\n            + [pl.col(c).cast(pl.Float32).quantile(0.75).alias(f'{c}_q75') \n               for c in all_cols]\n            # Range of values (max - min)\n            + [(pl.col(c).max() - pl.col(c).min()).cast(pl.Float32).alias(f'{c}_range') \n               for c in all_cols]\n        )\n        .collect()\n        .pivot('day_part', index='id', aggregate_function='mean', sort_columns=True)\n    )\n    return pl_df\n\ndef create_series_df(path):\n    '''\n    Read series files and aggregate in 2 steps to avoid out of memory \n    '''\n    series_pl_df_base = create_series_df_base(path)\n    day_part_agg_p1_df = day_4parts_agg_p1(series_pl_df_base)\n    day_part_agg_p2_df = day_4parts_agg_p2(series_pl_df_base)\n    series_pl_df = pl.concat(\n        [day_part_agg_p1_df, day_part_agg_p2_df], how='align'\n    )\n    return series_pl_df.to_pandas()","metadata":{"execution":{"iopub.status.busy":"2024-12-03T11:48:44.446546Z","iopub.execute_input":"2024-12-03T11:48:44.447571Z","iopub.status.idle":"2024-12-03T11:48:44.471238Z","shell.execute_reply.started":"2024-12-03T11:48:44.447516Z","shell.execute_reply":"2024-12-03T11:48:44.469758Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"#### Process data","metadata":{}},{"cell_type":"code","source":"%%time \n\n# Define file paths\nroot_path = '/kaggle/input/child-mind-institute-problematic-internet-use'\ntrain_series_path = f'{root_path}/series_train.parquet'\ntest_series_path = f'{root_path}/series_test.parquet'\n\nconcat_train_series_path = '/kaggle/working/series_df.parquet'\nconcat_test_series_path = '/kaggle/working/series_df_test.parquet'\n\n# Read base files\ndf = pd.read_csv(f'{root_path}/train.csv')\ndf_test = pd.read_csv(f'{root_path}/test.csv')\n\nsubmit = pd.read_csv(f'{root_path}/sample_submission.csv')\ndata_dict = pd.read_csv(f'{root_path}/data_dictionary.csv')\n\n# Aggregate series files and concatenate\nprocess = False\n\nif process:\n    series_df = create_series_df(train_series_path)\n    series_df.to_parquet('series_df.parquet')\n    series_df_test = create_series_df(test_series_path)\nelse:    \n    series_df = pd.read_parquet('/kaggle/input/cmi-df/series_df.parquet')\n    series_df_test = create_series_df(test_series_path)","metadata":{"_kg_hide-output":true,"execution":{"iopub.status.busy":"2024-12-03T11:48:44.473025Z","iopub.execute_input":"2024-12-03T11:48:44.473439Z","iopub.status.idle":"2024-12-03T11:48:45.169373Z","shell.execute_reply.started":"2024-12-03T11:48:44.473399Z","shell.execute_reply":"2024-12-03T11:48:45.168198Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Feature description query\ndata_dict.query(\"Field == 'PCIAT-PCIAT_Total'\")","metadata":{"execution":{"iopub.status.busy":"2024-12-03T11:48:45.172123Z","iopub.execute_input":"2024-12-03T11:48:45.172475Z","iopub.status.idle":"2024-12-03T11:48:45.202626Z","shell.execute_reply.started":"2024-12-03T11:48:45.172443Z","shell.execute_reply":"2024-12-03T11:48:45.201497Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"df.head()","metadata":{"execution":{"iopub.status.busy":"2024-12-03T11:48:45.204045Z","iopub.execute_input":"2024-12-03T11:48:45.204524Z","iopub.status.idle":"2024-12-03T11:48:45.235855Z","shell.execute_reply.started":"2024-12-03T11:48:45.204464Z","shell.execute_reply":"2024-12-03T11:48:45.23473Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"series_df.head()","metadata":{"execution":{"iopub.status.busy":"2024-12-03T11:48:45.237076Z","iopub.execute_input":"2024-12-03T11:48:45.237403Z","iopub.status.idle":"2024-12-03T11:48:45.263347Z","shell.execute_reply.started":"2024-12-03T11:48:45.237373Z","shell.execute_reply":"2024-12-03T11:48:45.262226Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"def build_autoencoder(input_dim, encoding_dim):\n    '''\n    Autoencoder to reduce dimensionality and break features' correlation\n    '''\n    input_layer = Input(shape=(input_dim,))\n    encoded = Dense(encoding_dim, activation='relu')(input_layer)\n    decoded = Dense(input_dim, activation='sigmoid')(encoded)\n    autoencoder = Model(inputs=input_layer, outputs=decoded)\n    encoder = Model(inputs=input_layer, outputs=encoded)\n    autoencoder.compile(optimizer=Adam(), loss='mse')\n    \n    return autoencoder, encoder\n\n# Encode aggregated series data\nencoding_dim = 50\n\nif encoding_dim != None:\n    series_df_enc = series_df.drop(columns='id')\n    \n    # Impute missing values\n    imputer = SimpleImputer(strategy='mean')\n    series_df_enc = imputer.fit_transform(series_df_enc)\n    \n    # Normalize data\n    q_transformer = QuantileTransformer(output_distribution='normal')\n    series_df_enc = q_transformer.fit_transform(series_df_enc)\n\n    # Construct and train the autoencoder\n    autoencoder, encoder = build_autoencoder(\n        input_dim=series_df_enc.shape[1], encoding_dim=encoding_dim,\n    )\n    autoencoder.fit(series_df_enc, series_df_enc, epochs=100, batch_size=32, \n                    shuffle=True, verbose=0)\n\n    # Encode train data and merge to the main df\n    series_df_enc = encoder.predict(series_df_enc)\n    series_df_enc = pd.DataFrame(\n        series_df_enc, columns=[f'Enc_{i+1}' for i in range(series_df_enc.shape[1])]\n    )\n    series_df_enc['id'] = series_df['id']\n    df = df.merge(series_df_enc, how='left', on='id')\n\n    # Apply the encodings on the test data and merge to the main df\n    series_df_test_enc = series_df_test.drop(columns='id')\n    series_df_test_enc = imputer.transform(series_df_test_enc)\n    series_df_test_enc = q_transformer.transform(series_df_test_enc)\n    series_df_test_enc = encoder.predict(series_df_test_enc)\n    series_df_test_enc = pd.DataFrame(\n        series_df_test_enc,\n        columns=[f'Enc_{i+1}' for i in range(series_df_test_enc.shape[1])]\n    )\n    series_df_test_enc['id'] = series_df_test['id']\n    df_test = df_test.merge(series_df_test_enc, how='left', on='id')\n    \nelse:\n    # Merge dataframes without encoding\n    df = df.merge(series_df, how='left', on='id')\n    df_test = df_test.merge(series_df_test, how='left', on='id')","metadata":{"execution":{"iopub.status.busy":"2024-12-03T11:48:45.264746Z","iopub.execute_input":"2024-12-03T11:48:45.26504Z","iopub.status.idle":"2024-12-03T11:49:03.248138Z","shell.execute_reply.started":"2024-12-03T11:48:45.265013Z","shell.execute_reply":"2024-12-03T11:49:03.246928Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"### Processing functions\n\ndef high_corr_cols(df, threshold):\n    '''\n    List of highly correlated columns with the highest total missing \n    values between correlated pairs\n    '''\n    corr_matrix = df.corr(numeric_only=True)\n    high_corr_to_drop = set()\n    \n    for i in range(len(corr_matrix.columns)):\n        for j in range(i+1, len(corr_matrix.columns)):\n            col1, col2 = corr_matrix.columns[i], corr_matrix.columns[j]\n            if abs(corr_matrix.loc[col1, col2]) > threshold:\n                if df[col1].isna().sum() >= df[col2].isna().sum():\n                    high_corr_to_drop.add(col1)\n                else:\n                    high_corr_to_drop.add(col2)\n                    \n    return list(high_corr_to_drop)\n\ndef encode_seasons(df):\n    '''\n    Based on feature importance analysis keep just the seasons\n    that have some importance to the models\n    '''\n    df['PAQ_C_winter'] = 0\n    df.loc[df['PAQ_C-Season'].eq('Winter'), 'PAQ_C_winter'] = 1\n\n    df['SDS_winter'] = 0\n    df.loc[df['SDS-Season'].eq('Winter'), 'SDS_winter'] = 1\n    df['SDS_spring'] = 0\n    df.loc[df['SDS-Season'].eq('Spring'), 'SDS_spring'] = 1\n\n    df['Physical_summer'] = 0\n    df.loc[df['Physical-Season'].eq('Summer'), 'Physical_summer'] = 1\n    df['Physical_fall'] = 0\n    df.loc[df['Physical-Season'].eq('Fall'), 'Physical_fall'] = 1\n\n    df['PAQ_A_summer'] = 0\n    df.loc[df['PAQ_A-Season'].eq('Summer'), 'PAQ_A_summer'] = 1\n\n    df['FGC_summer'] = 0\n    df.loc[df['FGC-Season'].eq('Summer'), 'FGC_summer'] = 1\n\n    df['Basic_summer'] = 0\n    df.loc[df['Basic_Demos-Enroll_Season'].eq('Summer'), 'Basic_summer'] = 1\n    df['Basic_fall'] = 0\n    df.loc[df['Basic_Demos-Enroll_Season'].eq('Fall'), 'Basic_fall'] = 1\n\n    df['Basic_summer'] = 0\n    df.loc[df['Fitness_Endurance-Season'].eq('Summer'), 'Basic_summer'] = 1\n\n    df['Basic_winter'] = 0\n    df.loc[df['PreInt_EduHx-Season'].eq('Winter'), 'Basic_winter'] = 1\n    \n    return df\n\ndef clean(df):\n    '''\n    Based on EDA drop erroneus values and unuseful features\n    '''\n    # Drop unuseful columns\n    cols_to_drop = ['BIA-BIA_FFM']\n    cols_to_drop += [col for col in X.columns if 'Zone' in col]\n    cols_to_drop += [col for col in X.columns if 'Season' in col]\n    df = df.drop(columns=cols_to_drop)\n    \n    # Change erroneus values with NaN\n    df.loc[df['Physical-BMI'].eq(0), 'Physical-BMI'] = np.nan\n    df.loc[df['Physical-Weight'].eq(0), 'Physical-Weight'] = np.nan\n    df.loc[df['FGC-FGC_GSND'].eq(0) | df['FGC-FGC_GSND'].gt(100), 'FGC-FGC_GSND'] = np.nan\n    df.loc[df['FGC-FGC_GSD'].eq(0) | df['FGC-FGC_GSD'].gt(100), 'FGC-FGC_GSD'] = np.nan\n    df.loc[df['FGC-FGC_SRR'].eq(0), 'FGC-FGC_SRR'] = np.nan\n    df.loc[df['FGC-FGC_TL'].eq(0), 'FGC-FGC_TL'] = np.nan\n    df.loc[df['BIA-BIA_BMC'].lt(0.1) | df['BIA-BIA_BMC'].gt(30), 'BIA-BIA_BMC'] = np.nan\n    df.loc[df['BIA-BIA_BMI'].lt(10), 'BIA-BIA_BMI'] = np.nan\n    df.loc[df['BIA-BIA_BMR'].gt(4000), 'BIA-BIA_BMR'] = np.nan\n    df.loc[df['BIA-BIA_DEE'].gt(8000), 'BIA-BIA_DEE'] = np.nan\n    df.loc[df['BIA-BIA_ECW'].gt(120), 'BIA-BIA_ECW'] = np.nan\n    df.loc[df['BIA-BIA_FFMI'].gt(90), 'BIA-BIA_FFMI'] = np.nan\n    df.loc[df['BIA-BIA_FMI'].le(0), 'BIA-BIA_FMI'] = np.nan\n    df.loc[df['BIA-BIA_Fat'].le(0), 'BIA-BIA_Fat'] = np.nan\n    df.loc[df['BIA-BIA_ICW'].gt(160), 'BIA-BIA_ICW'] = np.nan\n    df.loc[df['BIA-BIA_LDM'].gt(100), 'BIA-BIA_LDM'] = np.nan\n    df.loc[df['BIA-BIA_LST'].gt(400), 'BIA-BIA_LST'] = np.nan\n    df.loc[df['BIA-BIA_SMM'].gt(400), 'BIA-BIA_SMM'] = np.nan\n    df.loc[df['BIA-BIA_TBW'].gt(300), 'BIA-BIA_TBW'] = np.nan\n    df.loc[df['PAQ_A-PAQ_A_Total'].lt(0.5), 'PAQ_A-PAQ_A_Total'] = np.nan\n    df.loc[df['SDS-SDS_Total_Raw'].lt(20), 'SDS-SDS_Total_Raw'] = np.nan\n    df.loc[df['Fitness_Endurance-Max_Stage'].gt(15), 'Fitness_Endurance-Max_Stage'] = np.nan\n    \n    return df\n\ndef clust_feats(df, df_test, feats_cols):\n    '''\n    Add clustering features\n    '''\n    # Prepare data for unsupervised models\n    clust_df = df[feats_cols].select_dtypes(include=np.number)\n    clust_df_test = df_test[clust_df.columns]\n    \n    imputer = SimpleImputer(strategy='mean')\n    clust_df = imputer.fit_transform(clust_df)\n    clust_df_test = imputer.transform(clust_df_test)\n    \n    # Create unsupervised models\n    kmeans = KMeans(n_clusters=4)\n    gm = GaussianMixture(n_components=4)\n    \n    # Predict into new features\n    df['knn'] = kmeans.fit_predict(clust_df)\n    df['gauss_mix'] = gm.fit_predict(clust_df)\n    \n    df_test['knn'] = kmeans.predict(clust_df_test)\n    df_test['gauss_mix'] = gm.predict(clust_df_test)\n\n    # Convert to category dtype (will one hot encode them later)\n    clust_cols = ['knn', 'gauss_mix']\n    df[clust_cols] = df[clust_cols].astype('category')\n    df_test[clust_cols] = df_test[clust_cols].astype('category')\n\n    return df, df_test\n\ndef feat_eng(df):\n    '''\n    Feature engineering function\n    '''\n    df['bmi_age'] = df['Physical-BMI'] * df['Basic_Demos-Age']\n    df['inthour_age'] = df['PreInt_EduHx-computerinternet_hoursday'] * df['Basic_Demos-Age']\n    df['bmi_inthour'] = df['Physical-BMI'] * df['PreInt_EduHx-computerinternet_hoursday']\n    df['BFP_BMI'] = df['BIA-BIA_Fat'] / (df['BIA-BIA_BMI'] + 1e-5)\n    df['FFMI_BFP'] = df['BIA-BIA_FFMI'] / (df['BIA-BIA_Fat'] + 1e-5)\n    df['FMI_BFP'] = df['BIA-BIA_FMI'] / (df['BIA-BIA_Fat'] + 1e-5)\n    df['LST_TBW'] = df['BIA-BIA_LST'] / (df['BIA-BIA_TBW'] + 1e-5)\n    df['BFP_BMR'] = df['BIA-BIA_Fat'] * df['BIA-BIA_BMR']\n    df['BFP_DEE'] = df['BIA-BIA_Fat'] * df['BIA-BIA_DEE']\n    df['BMR_Weight'] = df['BIA-BIA_BMR'] / (df['Physical-Weight'] + 1e-5)\n    df['DEE_Weight'] = df['BIA-BIA_DEE'] / (df['Physical-Weight'] + 1e-5)\n    df['SMM_Height'] = df['BIA-BIA_SMM'] / (df['Physical-Height'] + 1e-5)\n    df['Muscle_to_Fat'] = df['BIA-BIA_SMM'] / (df['BIA-BIA_FMI'] + 1e-5)\n    df['Hydration_Status'] = df['BIA-BIA_TBW'] / (df['Physical-Weight'] + 1e-5)\n    df['ICW_TBW'] = df['BIA-BIA_ICW'] / (df['BIA-BIA_TBW'] + 1e-5)\n    \n    return df","metadata":{"execution":{"iopub.status.busy":"2024-12-03T11:49:03.249759Z","iopub.execute_input":"2024-12-03T11:49:03.250096Z","iopub.status.idle":"2024-12-03T11:49:03.277623Z","shell.execute_reply.started":"2024-12-03T11:49:03.250062Z","shell.execute_reply":"2024-12-03T11:49:03.276288Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"Train data has samples with missing target values. Before droping these samples, we will make use of them by training 2 unsupervised models to cluster the samples into 4 groups and add these predictions as additional features to the train set. We don't expect that these predictions will correspond to our target classes, but at least we extract some clustering information from the entire available data and hope that these separation into groups will help our final models.","metadata":{}},{"cell_type":"code","source":"# Add clustering features\nfeats_cols = df_test.columns.to_list()\nfeats_cols.remove('id')\ndf, df_test = clust_feats(df, df_test, feats_cols)\n\n# Separate all the targets\nall_target_cols = [col for col in df if col not in df_test]\ny_full = df[all_target_cols]\n\n# Indexes where number of missing tests is less than 2\nmulti_target_cols = [col for col in all_target_cols if col not in \n                     ['sii', 'PCIAT-PCIAT_Total', 'PCIAT-Season']]\nkeep_idx = y_full[multi_target_cols].isna().sum(axis=1).lt(2)\n\n# Multi class target\ny = y_full.loc[keep_idx, 'sii']\n\n# Regression target as sum of all tests \ny_total = y_full.loc[keep_idx, 'PCIAT-PCIAT_Total']\n\n# Multi output multi class target\ny_multi = y_full.loc[keep_idx, multi_target_cols]\ny_multi = y_multi.fillna(0)\n\n# Train / test sets\nX_test = df_test.drop(columns='id')\nX = df.loc[keep_idx, X_test.columns]\n\n# Preprocess\nX = feat_eng(X)\nX_test = feat_eng(X_test)\n\nX = encode_seasons(X)\nX_test = encode_seasons(X_test)\n\nX = clean(X)\nX_test = clean(X_test)\n\n# Drop highly correlated columns\nhigh_corr_cols_list = high_corr_cols(X, threshold=0.90)\nX = X.drop(columns=high_corr_cols_list)\nX_test = X_test.drop(columns=high_corr_cols_list)\n\n# Columns types\nnum_cols = X.select_dtypes(np.number).columns\ncat_cols = X.select_dtypes(include=['object', 'category']).columns\n\n# Pipeline to process numerical columns\nnum_pipeline = make_pipeline(\n    QuantileTransformer(output_distribution='normal'),\n    )\n# Pipeline to process categorical columns\ncat_pipeline = make_pipeline(\n    OneHotEncoder(\n        dtype=np.int8, drop='if_binary', sparse_output=False,\n        min_frequency=0.03, handle_unknown='infrequent_if_exist'),\n    )\n# Define column transformer\nprocessor = make_column_transformer(\n    (num_pipeline, num_cols),\n    (cat_pipeline, cat_cols),\n    verbose_feature_names_out=False,\n    verbose=False,\n    ).set_output(transform='pandas')","metadata":{"execution":{"iopub.status.busy":"2024-12-03T12:36:13.708117Z","iopub.execute_input":"2024-12-03T12:36:13.708609Z","iopub.status.idle":"2024-12-03T12:36:16.268036Z","shell.execute_reply.started":"2024-12-03T12:36:13.708573Z","shell.execute_reply":"2024-12-03T12:36:16.266774Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Dropped highly correlated columns\nprint(high_corr_cols_list)","metadata":{"execution":{"iopub.status.busy":"2024-12-03T11:49:08.897146Z","iopub.execute_input":"2024-12-03T11:49:08.897507Z","iopub.status.idle":"2024-12-03T11:49:08.903425Z","shell.execute_reply.started":"2024-12-03T11:49:08.897475Z","shell.execute_reply":"2024-12-03T11:49:08.90217Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"### Plotting functions\n\ndef plot_corr_to_target(corr, target):\n    '''\n    Plot the correlation of all the features related to target\n    '''\n    corr_to_target = corr[target].sort_values(ascending=False)\n    corr_to_target[1:].plot(kind='barh', figsize=(7, 10), grid=True,\n                            title=f'Features correlation to {target}',\n                            )\n    plt.show()\n\ndef plot_feats_corr(corr):\n    '''\n    Plot the correlation between the features\n    '''\n    top_right_mask = np.triu(np.ones_like(corr, dtype=bool))\n    plt.figure(figsize=(18, 14))\n    sns.heatmap(\n        corr,\n        # mask=(corr<0.9),\n        # mask=top_right_mask,\n        linewidth=0.5,\n        linecolor='white',\n        annot=True,\n        annot_kws={'size': 7},\n        cmap='crest',\n    )\n    plt.title('Correlation between features')\n    plt.tight_layout()\n    plt.show()\n\ndef detect_outliers(df, column):\n    Q1 = df[column].quantile(0.25)\n    Q3 = df[column].quantile(0.75)\n    IQR = Q3 - Q1\n    lower_bound = Q1 - 1.5 * IQR\n    upper_bound = Q3 + 1.5 * IQR\n\n    return df[df[column].lt(lower_bound) | df[column].gt(upper_bound)]\n\ndef scatterplot_all(df, target):\n    '''\n    Scatterplot all the features outliers against the target variable\n    '''\n    features = df.select_dtypes(np.number).columns\n    num_cols = 4\n    num_rows = (len(features) + num_cols - 1) // num_cols\n\n    plt.figure(figsize=(num_cols * 4, num_rows * 2)) \n\n    for i, feature in enumerate(features):\n        plt.subplot(num_rows, num_cols, i + 1)\n        sns.scatterplot(data=df, x=feature, y=target, \n                        color='blue', label='Data', alpha=0.1)\n        \n        outliers = detect_outliers(df, feature)\n        sns.scatterplot(data=outliers, x=feature, y=target, \n                        color='red', label='Outliers', alpha=0.1)\n        \n        plt.title(f'{feature} vs {target}')\n        plt.xlabel(feature)\n        plt.ylabel(target)\n\n        plt.legend(loc='upper right')\n\n    plt.tight_layout()\n    plt.show()\n\ndef boxplot_all(df):\n    '''\n    Boxplot all the features outliers\n    '''\n    features = df.select_dtypes(np.number).columns\n    num_cols = 3\n    num_rows = (len(features) + num_cols - 1) // num_cols\n\n    plt.figure(figsize=(num_cols * 5, num_rows * 1))\n\n    for i, feature in enumerate(features):\n        plt.subplot(num_rows, num_cols, i + 1)\n        sns.boxplot(data=df, x=feature, color='blue')\n        plt.title(f'{feature} outliers')\n        plt.xlabel(feature)\n\n    plt.tight_layout()\n    plt.show()\n\ndef histplot_all(df, target):\n    '''\n    Plot all the values distributions\n    '''\n    features = df.columns\n    num_cols = 4\n    num_rows = (len(features) + num_cols - 1) // num_cols\n\n    plt.figure(figsize=(num_cols * 4, num_rows * 4))\n\n    for i, feature in enumerate(features):\n        plt.subplot(num_rows, num_cols, i + 1)\n        sns.histplot(df, x=feature, bins=30, hue=target)\n        plt.title(f'Distribution of {feature}')\n        plt.xticks(rotation=45, ha='right')\n        plt.xlabel(feature)\n        plt.ylabel('Frequency')\n        \n    plt.tight_layout()\n    plt.show()","metadata":{"execution":{"iopub.status.busy":"2024-12-03T11:49:08.905618Z","iopub.execute_input":"2024-12-03T11:49:08.906065Z","iopub.status.idle":"2024-12-03T11:49:08.926525Z","shell.execute_reply.started":"2024-12-03T11:49:08.90602Z","shell.execute_reply":"2024-12-03T11:49:08.92553Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"plot = True\n\n# Plot data for analysis\nif plot:\n    plot_cols = [col for col in X.columns if 'Enc' not in col]\n    df_plot = X[plot_cols].join(y)\n    target = 'sii'\n    corr = df_plot.corr(numeric_only=True)\n    plot_corr_to_target(corr, target)\n    plot_feats_corr(round(corr*100))\n    scatterplot_all(df_plot, target)\n    boxplot_all(df_plot)\n    histplot_all(df_plot, target)","metadata":{"execution":{"iopub.status.busy":"2024-12-03T11:49:08.928008Z","iopub.execute_input":"2024-12-03T11:49:08.929307Z","iopub.status.idle":"2024-12-03T11:49:50.295466Z","shell.execute_reply.started":"2024-12-03T11:49:08.929258Z","shell.execute_reply":"2024-12-03T11:49:50.294465Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"#### Optuna tuning","metadata":{}},{"cell_type":"code","source":"def qwk_on_sii(y_true, y_pred):\n    '''\n    Metric used for evaluation\n    '''\n    y_pred_int = y_pred.clip(0, 3).round().astype(int)\n    \n    return cohen_kappa_score(y_true, y_pred_int, weights='quadratic')\n\n# Use gpu if available\ndevice = 'gpu' if torch.cuda.is_available() else 'cpu'\n\ndef model_objective(trial, model_type):\n    \"\"\"\n    Objective function for hyperparameter tuning using Optuna \n    \"\"\"\n    if model_type == 'lgb':\n        params = {\n            'n_estimators': 2000,\n            'num_leaves': trial.suggest_int('num_leaves', 2, 250),\n            'min_child_samples': trial.suggest_int('min_child_samples', 5, 1000),\n            'learning_rate': trial.suggest_uniform('learning_rate', 0.005, 0.1),\n            'reg_alpha': trial.suggest_loguniform('reg_alpha', 1e-8, 10),\n            'reg_lambda': trial.suggest_loguniform('reg_lambda', 1e-8, 10),\n            'colsample_bytree': trial.suggest_uniform('colsample_bytree', 0.5, 1),\n            'subsample': trial.suggest_uniform('subsample', 0.5, 1),\n            'subsample_freq': trial.suggest_int('subsample_freq', 0, 10),\n            \n            'objective': 'regression',\n            'device': device,\n            'verbosity': -1,\n        }\n        model = lgb.LGBMRegressor(**params)\n\n    elif model_type == 'xgb':\n        params = {\n            'n_estimators': 2000,\n            'learning_rate': trial.suggest_uniform('learning_rate', 0.005, 0.1),\n            'max_depth': trial.suggest_int('max_depth', 4, 20),\n            'min_child_weight': trial.suggest_int('min_child_weight', 1, 1000),\n            'max_delta_step': trial.suggest_int('max_delta_step', 0, 10),\n            'subsample': trial.suggest_uniform('subsample', 0.5, 1),\n            'colsample_bytree': trial.suggest_uniform('colsample_bytree', 0.5, 1),\n            'lambda': trial.suggest_loguniform('reg_lambda', 1e-8, 10),\n            'alpha': trial.suggest_loguniform('reg_alpha', 1e-8, 10),\n            'gamma': trial.suggest_loguniform('gamma', 1e-8, 10),\n            \n            'objective': 'reg:squarederror',\n            'device': device,\n            'tree_method': 'hist',\n            'verbosity': 0,\n        }\n        model = xgb.XGBRegressor(**params, early_stopping_rounds=50)\n    \n    elif model_type == 'cb':\n        params = {\n            'iterations': 2000,\n            'learning_rate': trial.suggest_float('learning_rate', 0.001, 0.1, log=True),\n            'depth': trial.suggest_int('depth', 4, 10),\n            'l2_leaf_reg': trial.suggest_float('l2_leaf_reg', 0.00000001, 100.0, log=True),\n            'random_strength': trial.suggest_float('random_strength', 0.00000001, 10.0, log=True),\n            'boosting_type': trial.suggest_categorical('boosting_type', ['Ordered', 'Plain']),\n            'bagging_temperature': trial.suggest_float('bagging_temperature', 0, 10),\n            \n            'task_type': device.upper(),\n            'verbose': True,\n        }\n        model = cb.CatBoostRegressor(**params, early_stopping_rounds=50)\n        \n    else:\n        raise ValueError(\"Supported model types: 'lgb', 'xgb', 'cb'\")\n\n    # Split data in folds\n    skf = StratifiedKFold(n_splits=3, shuffle=True)\n    cv_results = []\n\n    for train_idx, val_idx in skf.split(X, y):\n        X_train, X_val = X.iloc[train_idx], X.iloc[val_idx]\n        y_train, y_val = y.iloc[train_idx], y.iloc[val_idx]\n        \n        # Transform data for the current fold\n        X_train_trans = processor.fit_transform(X_train, y_train)\n        X_val_trans = processor.transform(X_val)\n        \n        # Train model\n        if model_type == 'lgb':\n            model.fit(X_train_trans, y_train,\n                      eval_set=[(X_val_trans, y_val)],\n                      callbacks=[lgb.early_stopping(50)],\n                     )\n        else:\n            model.fit(X_train_trans, y_train,\n                      eval_set=[(X_val_trans, y_val)],\n                     )\n        # Predict and accumulate results from every fold\n        y_pred = model.predict(X_val_trans)\n        cv_results.append(qwk_on_sii(y_val, y_pred))\n\n    return np.mean(cv_results)","metadata":{"execution":{"iopub.status.busy":"2024-12-03T11:49:50.297139Z","iopub.execute_input":"2024-12-03T11:49:50.297608Z","iopub.status.idle":"2024-12-03T11:49:50.323708Z","shell.execute_reply.started":"2024-12-03T11:49:50.297561Z","shell.execute_reply":"2024-12-03T11:49:50.322378Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"%%time\nuse_optuna = False\n\nif use_optuna:\n    # Optuna study\n    def objective(trial):\n        return model_objective(trial, model_type='cb')\n        \n    sampler = optuna.samplers.TPESampler(multivariate=True)\n    study = optuna.create_study(direction='maximize', sampler=sampler)\n    study.optimize(objective, n_trials=50, show_progress_bar=True)\n\n    # Show best results\n    trial = study.best_trial\n\n    print('Number of finished trials: ', len(study.trials))\n    print('Best trial:')\n    print('Value:', trial.value)\n    print('Params:')\n\n    for key, value in trial.params.items():\n        print(f'{key}: {value}')","metadata":{"execution":{"iopub.status.busy":"2024-12-03T11:49:50.325198Z","iopub.execute_input":"2024-12-03T11:49:50.325641Z","iopub.status.idle":"2024-12-03T11:49:50.337457Z","shell.execute_reply.started":"2024-12-03T11:49:50.325595Z","shell.execute_reply":"2024-12-03T11:49:50.3363Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Best LGB parameters\nlgb_params = {\n    'n_estimators': 250,\n    \n    'num_leaves': 11, \n    'min_child_samples': 17, \n    'learning_rate': 0.02201701992617784, \n    'reg_alpha': 0.0024384396081629394, \n    'reg_lambda': 3.0695161823482566, \n    'colsample_bytree': 0.6616740997193433, \n    'subsample': 0.7321840655648034, \n    'subsample_freq': 9,\n    \n    'device': device,\n    'objective': 'regression',\n    'verbosity': -1,\n    'n_jobs': -1,\n}\n# Best XGB parameters\nxgb_params = {\n    'n_estimators': 200,\n    \n    'learning_rate': 0.05392429821566442, \n    'max_depth': 10, \n    'min_child_weight': 62, \n    'max_delta_step': 0, \n    'subsample': 0.5690276490941832, \n    'colsample_bytree': 0.7712620459922648, \n    'reg_lambda': 8.327490867779473e-06, \n    'reg_alpha': 4.334646806827537, \n    'gamma': 0.0007750927775908343,\n    \n    'device': device,\n    'verbosity': 0,\n    'n_jobs': -1,\n}\n# CatBoostClassifier parameters \ncb_params = {\n    'iterations': 150,\n    \n    'learning_rate': 0.08025400759942777, \n    'depth': 4, \n    'l2_leaf_reg': 5.0402147152584106e-08, \n    'random_strength': 2.6263000467118993, \n    'boosting_type': 'Plain', \n    'bagging_temperature': 0.9738097766948606,\n    \n    'task_type': device.upper(),\n    'verbose': False,\n}","metadata":{"execution":{"iopub.status.busy":"2024-12-03T11:49:50.338898Z","iopub.execute_input":"2024-12-03T11:49:50.339346Z","iopub.status.idle":"2024-12-03T11:49:50.35524Z","shell.execute_reply.started":"2024-12-03T11:49:50.3393Z","shell.execute_reply":"2024-12-03T11:49:50.354125Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"#### Test models","metadata":{}},{"cell_type":"code","source":"def create_model(model_type):\n    '''\n    Create a model with it's tuned parameters\n    '''\n    if model_type == 'lgb':\n        model = lgb.LGBMRegressor(**lgb_params)\n    \n    elif model_type == 'xgb':\n        model = xgb.XGBRegressor(**xgb_params)\n    \n    elif model_type == 'cb':\n        model = cb.CatBoostRegressor(**cb_params)\n        \n    elif model_type == 'multi_lgb':\n        model = MultiOutputRegressor(lgb.LGBMRegressor(**lgb_params))\n    \n    elif model_type == 'multi_xgb':\n        model = MultiOutputRegressor(xgb.XGBRegressor(**xgb_params))\n        \n    elif model_type == 'multi_cb':\n        model = MultiOutputRegressor(cb.CatBoostRegressor(**cb_params))\n        \n    else:\n        raise ValueError(\"Model type not supported\")\n        \n    return model\n\ndef y_total_to_class(y_total):\n    '''\n    Transform continous value target to it's corresponding class\n    '''\n    y_int = y_total.copy()\n    y_int[y_int < 30.5] = 0\n    y_int[(y_int >= 30.5) & (y_int < 49.5)] = 1\n    y_int[(y_int >= 49.5) & (y_int < 79.5)] = 2\n    y_int[y_int >= 79.5] = 3\n    \n    return y_int.astype(int)\n\ndef tune_predictions(y_pred_float, thresholds):\n    '''\n    Separate into classes based on tuned thresholds\n    '''\n    thresholds = sorted(thresholds)\n    \n    return pd.cut(y_pred_float, [-np.inf] + thresholds + [np.inf], \n                  labels=[0, 1, 2, 3])\n\ndef evaluate_predictions(thresholds, y_true, y_pred_float):\n    '''\n    Metric used for evaluation of tuned_predictions\n    '''\n    y_pred_int = tune_predictions(y_pred_float, thresholds)\n    \n    return -cohen_kappa_score(y_true, y_pred_int, weights = 'quadratic')\n\ndef plot_conf_matrix(y_true, y_pred):\n    '''\n    Plot confusion matrix\n    '''\n    conf_matrix = confusion_matrix(y_true, y_pred)\n    plt.figure(figsize=(3, 2))\n    sns.heatmap(conf_matrix, annot=True, fmt='d', cmap='summer', \n                cbar=False, linewidths=0.5)\n    plt.title('Confusion Matrix')\n    plt.xlabel('Predicted')\n    plt.ylabel('True')\n    plt.show()","metadata":{"execution":{"iopub.status.busy":"2024-12-03T11:49:50.356445Z","iopub.execute_input":"2024-12-03T11:49:50.356756Z","iopub.status.idle":"2024-12-03T11:49:50.370436Z","shell.execute_reply.started":"2024-12-03T11:49:50.356727Z","shell.execute_reply":"2024-12-03T11:49:50.369242Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"%%time\n\n### Group 1: 3 single regressors predicting 'sii' as target\n\n# Create training pipeline\nlgb_model = create_model('lgb')\nlgb_pipeline = make_pipeline(processor, lgb_model)\nlgb_pred = cross_val_predict(lgb_pipeline, X, y, cv=5)\n\nxgb_model = create_model('xgb')\nxgb_pipeline = make_pipeline(processor, xgb_model)\nxgb_pred = cross_val_predict(xgb_pipeline, X, y, cv=5)\n\ncb_model = create_model('cb')\ncb_pipeline = make_pipeline(processor, cb_model)\ncb_pred = cross_val_predict(cb_pipeline, X, y, cv=5)\n\nmean_pred = np.mean([lgb_pred.clip(0, 3), \n                     xgb_pred.clip(0, 3), \n                     cb_pred.clip(0, 3)], axis=0)\nqwk_score = qwk_on_sii(y, mean_pred)\nprint(f'Group 1, mean QWK: {qwk_score.round(3)}')\n\nplot_conf_matrix(y, mean_pred.round())","metadata":{"execution":{"iopub.status.busy":"2024-12-03T11:49:50.371691Z","iopub.execute_input":"2024-12-03T11:49:50.372018Z","iopub.status.idle":"2024-12-03T11:50:37.582399Z","shell.execute_reply.started":"2024-12-03T11:49:50.371988Z","shell.execute_reply":"2024-12-03T11:50:37.581282Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Optimize thresholds\ngr1_optimizer = minimize(evaluate_predictions, x0=[0.5, 1.0, 1.5],\n                     args=(y, mean_pred), method='Nelder-Mead')\ngr1_pred_tuned = tune_predictions(mean_pred, gr1_optimizer.x)\n\nqwk_score_tuned = cohen_kappa_score(y, gr1_pred_tuned, weights = 'quadratic')\nprint(f'Group 1, tuned QWK: {qwk_score_tuned.round(3)}')\nprint(f'Optimized thresholds: {gr1_optimizer.x}')\n\nplot_conf_matrix(y, gr1_pred_tuned)","metadata":{"execution":{"iopub.status.busy":"2024-12-03T11:50:37.583895Z","iopub.execute_input":"2024-12-03T11:50:37.584353Z","iopub.status.idle":"2024-12-03T11:50:38.052856Z","shell.execute_reply.started":"2024-12-03T11:50:37.584306Z","shell.execute_reply":"2024-12-03T11:50:38.051802Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"%%time\n\n### Group 2: 3 multi output regressors predicting 20 PCIAT tests as targets\n\n# Train and predict with each model\nmulti_lgb = create_model('multi_lgb')\nlgb_train_pipeline = make_pipeline(processor, multi_lgb)\nlgb_pred = cross_val_predict(lgb_train_pipeline, X, y_multi, cv=5)\nlgb_pred_round_sum = lgb_pred.clip(0, 5).round().astype(int).sum(axis=1)\n\nmulti_xgb = create_model('multi_xgb')\nxgb_train_pipeline = make_pipeline(processor, multi_xgb)\nxgb_pred = cross_val_predict(xgb_train_pipeline, X, y_multi, cv=5)\nxgb_pred_round_sum = xgb_pred.clip(0, 5).round().astype(int).sum(axis=1)\n\nmulti_cb = create_model('multi_cb')\ncb_train_pipeline = make_pipeline(processor, multi_cb)\ncb_pred = cross_val_predict(cb_train_pipeline, X, y_multi, cv=5)\ncb_pred_round_sum = cb_pred.clip(0, 5).round().astype(int).sum(axis=1)\n\n# Average of sums of 20 predictions\npred_round_mean = np.mean([lgb_pred_round_sum, \n                          xgb_pred_round_sum, \n                          cb_pred_round_sum], axis=0).round()\npred_class = y_total_to_class(pred_round_mean)\nqwk_score = cohen_kappa_score(y, pred_class, weights='quadratic')\nprint(f'Group 2, mean QWK: {qwk_score.round(3)}')\n\nplot_conf_matrix(y, pred_class)","metadata":{"execution":{"iopub.status.busy":"2024-12-03T11:50:38.054033Z","iopub.execute_input":"2024-12-03T11:50:38.05495Z","iopub.status.idle":"2024-12-03T12:03:10.986337Z","shell.execute_reply.started":"2024-12-03T11:50:38.0549Z","shell.execute_reply":"2024-12-03T12:03:10.985274Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Sum of uncliped, non-rounded predictions\nlgb_pred_sum = lgb_pred.sum(axis=1)\nxgb_pred_sum = xgb_pred.sum(axis=1)\ncb_pred_sum = cb_pred.sum(axis=1)\n\n# Optimize thresholds\npred_mean = np.mean([lgb_pred_sum, xgb_pred_sum, cb_pred_sum], axis=0)\n\ngr2_optimizer = minimize(evaluate_predictions, x0=[30.0, 40.0, 55.0],\n                     args=(y, pred_mean), method='Nelder-Mead')\ngr2_pred_tuned = tune_predictions(pred_mean, gr2_optimizer.x)\n\nqwk_score_tuned = cohen_kappa_score(y, gr2_pred_tuned, weights = 'quadratic')\nprint(f'Group 2 , tuned QWK: {qwk_score_tuned.round(3)}')\nprint(f'Optimized thresholds: {gr2_optimizer.x}')\n\nplot_conf_matrix(y, gr2_pred_tuned)","metadata":{"execution":{"iopub.status.busy":"2024-12-03T12:03:10.987689Z","iopub.execute_input":"2024-12-03T12:03:10.988129Z","iopub.status.idle":"2024-12-03T12:03:11.562772Z","shell.execute_reply.started":"2024-12-03T12:03:10.988081Z","shell.execute_reply":"2024-12-03T12:03:11.561797Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"### Group 3: 3 single regressors trained on meta features\n\n# Create meta features by stacking together Group 2 predictions \nX_meta = np.hstack([lgb_pred, xgb_pred, cb_pred])\n\n# Train and predict on meta features\nlgb_lv2 = create_model('lgb')\nlgb_pred_lv2 = cross_val_predict(lgb_lv2, X_meta, y, cv=5)\n\nxgb_lv2 = create_model('xgb')\nxgb_pred_lv2 = cross_val_predict(xgb_lv2, X_meta, y, cv=5)\n\ncb_lv2 = create_model('cb')\ncb_pred_lv2 = cross_val_predict(cb_lv2, X_meta, y, cv=5)\n\nmean_pred = np.mean([lgb_pred_lv2.clip(0, 3), \n                     xgb_pred_lv2.clip(0, 3), \n                     cb_pred_lv2.clip(0, 3)], axis=0)\nqwk_score = qwk_on_sii(y, pred_class)\nprint(f'Group 3, mean QWK: {qwk_score.round(3)}')\n\nplot_conf_matrix(y, mean_pred.round())","metadata":{"execution":{"iopub.status.busy":"2024-12-03T12:03:11.564009Z","iopub.execute_input":"2024-12-03T12:03:11.564452Z","iopub.status.idle":"2024-12-03T12:03:22.012834Z","shell.execute_reply.started":"2024-12-03T12:03:11.564407Z","shell.execute_reply":"2024-12-03T12:03:22.011719Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Optimize thresholds\ngr3_optimizer = minimize(evaluate_predictions, x0=[0.5, 1.0, 1.5], \n                 args=(y, mean_pred), method='Nelder-Mead')\ngr3_pred_tuned = tune_predictions(mean_pred, gr3_optimizer.x)\n\nqwk_score_tuned = cohen_kappa_score(y, gr3_pred_tuned, weights = 'quadratic')\nprint(f'Group 3, tuned QWK: {qwk_score_tuned.round(3)}')\nprint(f'Optimized thresholds: {gr3_optimizer.x}')\n\nplot_conf_matrix(y, gr3_pred_tuned)","metadata":{"execution":{"iopub.status.busy":"2024-12-03T12:03:22.014043Z","iopub.execute_input":"2024-12-03T12:03:22.014475Z","iopub.status.idle":"2024-12-03T12:03:22.473707Z","shell.execute_reply.started":"2024-12-03T12:03:22.01443Z","shell.execute_reply":"2024-12-03T12:03:22.472594Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"### Average of predictions from all 3 groups\npred_fin = np.mean([gr1_pred_tuned.astype(int), \n                    gr2_pred_tuned.astype(int), \n                    gr3_pred_tuned.astype(int)], axis=0).round().astype(int)\nqwk_fin = cohen_kappa_score(y, pred_fin, weights = 'quadratic')\nprint(f'Final cross-validated QWK: {qwk_fin.round(3)}')\n\nplot_conf_matrix(y, pred_fin)","metadata":{"execution":{"iopub.status.busy":"2024-12-03T12:03:22.474905Z","iopub.execute_input":"2024-12-03T12:03:22.475359Z","iopub.status.idle":"2024-12-03T12:03:22.636631Z","shell.execute_reply.started":"2024-12-03T12:03:22.475313Z","shell.execute_reply":"2024-12-03T12:03:22.635603Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"The 3rd class is very rare in the training set and we observe that we were able to predict samples into this class just after the threshold optimization.","metadata":{}},{"cell_type":"markdown","source":"#### Train and predict for submission","metadata":{}},{"cell_type":"code","source":"# Process data\nX_trans = pd.DataFrame(\n    processor.fit_transform(X, y), \n    columns=processor.get_feature_names_out(),\n    index=X.index\n    )\nX_test_trans = pd.DataFrame(\n    processor.transform(X_test), \n    columns=processor.get_feature_names_out(),\n    index=X_test.index\n    )\nprint(f'Nr of features in training set: {X_trans.shape[1]}')","metadata":{"execution":{"iopub.status.busy":"2024-12-03T12:36:21.62425Z","iopub.execute_input":"2024-12-03T12:36:21.624658Z","iopub.status.idle":"2024-12-03T12:36:22.514392Z","shell.execute_reply.started":"2024-12-03T12:36:21.624621Z","shell.execute_reply":"2024-12-03T12:36:22.513196Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"%%time\n\ndef predict_with(model_type, X_trans, y_final, X_test_trans):\n    X_shuffled, y_shuffled = shuffle(X_trans, y_final)\n    model = create_model(model_type)\n    model.fit(X_shuffled, y_shuffled)\n\n    return model.predict(X_test_trans)\n\n### Group 1 \n\nlgb_pred = predict_with('lgb', X_trans, y, X_test_trans)\nxgb_pred = predict_with('xgb', X_trans, y, X_test_trans)\ncb_pred = predict_with('cb', X_trans, y, X_test_trans)\n\nmean_pred = np.mean([lgb_pred.clip(0, 3), \n                     xgb_pred.clip(0, 3), \n                     cb_pred.clip(0, 3)], axis=0)    \ngr1_pred = tune_predictions(mean_pred, gr1_optimizer.x)\n\n### Group 2\n\n# Predict X_test's 20 PCIA targets \nlgb_pred = predict_with('multi_lgb', X_trans, y_multi, X_test_trans)\nlgb_pred_sum = lgb_pred.sum(axis=1)\n\nxgb_pred = predict_with('multi_xgb', X_trans, y_multi, X_test_trans)\nxgb_pred_sum = xgb_pred.sum(axis=1)\n\ncb_pred = predict_with('multi_cb', X_trans, y_multi, X_test_trans)\ncb_pred_sum = cb_pred.sum(axis=1)\n\npred_sum = np.mean([lgb_pred_sum, xgb_pred_sum, cb_pred_sum], axis=0)\ngr2_pred = tune_predictions(pred_sum, gr2_optimizer.x)\n\n### Group 3\n\n# On the 1st level predict 20 PCIA targets from the train set, to use them \n# as meta features for training the 2nd level estimator\nlgb_model = create_model('multi_lgb')\nlgb_train_pipeline = make_pipeline(processor, lgb_model)\nlgb_X_meta = cross_val_predict(lgb_train_pipeline, X, y_multi, cv=5)\n\nxgb_model = create_model('multi_xgb')\nxgb_train_pipeline = make_pipeline(processor, xgb_model)\nxgb_X_meta = cross_val_predict(xgb_train_pipeline, X, y_multi, cv=5)\n\ncb_model = create_model('multi_cb')\ncb_train_pipeline = make_pipeline(processor, cb_model)\ncb_X_meta = cross_val_predict(cb_train_pipeline, X, y_multi, cv=5)\n\n# Stack train meta features\nX_meta = np.hstack([lgb_X_meta, xgb_X_meta, cb_X_meta])\n\n# Predict meta features for the test set\nlgb_X_test_meta = predict_with('multi_lgb', X_trans, y_multi, X_test_trans)\nlgb_X_test_meta = lgb_X_test_meta\n\nxgb_X_test_meta = predict_with('multi_xgb', X_trans, y_multi, X_test_trans)\nxgb_X_test_meta = xgb_X_test_meta\n\ncb_X_test_meta = predict_with('multi_cb', X_trans, y_multi, X_test_trans)\ncb_X_test_meta = cb_X_test_meta\n\n# Stack test meta features\nX_test_meta = np.hstack([lgb_X_test_meta, xgb_X_test_meta, cb_X_test_meta])\n\n# Train and predict on meta features\nlgb_pred = predict_with('lgb', X_meta, y, X_test_meta)\nxgb_pred = predict_with('xgb', X_meta, y, X_test_meta)\ncb_pred = predict_with('cb', X_meta, y, X_test_meta)\n\nmean_pred = np.mean([lgb_pred.clip(0, 3), \n                     xgb_pred.clip(0, 3), \n                     cb_pred.clip(0, 3)], axis=0)    \ngr3_pred = tune_predictions(mean_pred, gr3_optimizer.x)\n\n### Final prediction\ny_pred = np.mean([gr1_pred.astype(int), \n                  gr2_pred.astype(int), \n                  gr3_pred.astype(int)], axis=0).round().astype(int)","metadata":{"execution":{"iopub.status.busy":"2024-12-03T12:12:19.233254Z","iopub.execute_input":"2024-12-03T12:12:19.233847Z","iopub.status.idle":"2024-12-03T12:30:57.611171Z","shell.execute_reply.started":"2024-12-03T12:12:19.23379Z","shell.execute_reply":"2024-12-03T12:30:57.609979Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Submission\nsubmit.sii = y_pred\nsubmit.to_csv('submission.csv', index=False)\nsubmit.head()","metadata":{"execution":{"iopub.status.busy":"2024-12-03T12:30:57.613739Z","iopub.execute_input":"2024-12-03T12:30:57.614117Z","iopub.status.idle":"2024-12-03T12:30:57.629339Z","shell.execute_reply.started":"2024-12-03T12:30:57.614082Z","shell.execute_reply":"2024-12-03T12:30:57.628137Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"plot_importance = True\n\nif plot_importance:\n    # Plot LGB model features importance\n    model = create_model('lgb')\n    model.fit(X_trans, y)\n    fig, ax = plt.subplots(figsize=(8, 50))\n    lgb.plot_importance(\n        model, \n        importance_type='gain', \n        max_num_features=None, \n        height=0.5,\n        grid=False,\n        precision=0,\n        title='LGBM feature importance',\n        ax=ax,\n        ignore_zero=True,\n        )\n    plt.show()","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2024-12-03T12:30:57.63071Z","iopub.execute_input":"2024-12-03T12:30:57.631049Z","iopub.status.idle":"2024-12-03T12:31:03.284274Z","shell.execute_reply.started":"2024-12-03T12:30:57.631017Z","shell.execute_reply":"2024-12-03T12:31:03.282925Z"}},"outputs":[],"execution_count":null}]}