{"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":"none","dataSources":[{"sourceId":50160,"databundleVersionId":7921029,"sourceType":"competition"}],"dockerImageVersionId":30698,"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":false}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"# Import necessary libraries\nimport polars as pl\nimport numpy as np\nimport pandas as pd\nimport lightgbm as lgb\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.metrics import roc_auc_score\nfrom sklearn.preprocessing import StandardScaler\nfrom scipy.stats import norm\n\n\n\n","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2024-05-14T09:13:22.086567Z","iopub.execute_input":"2024-05-14T09:13:22.087752Z","iopub.status.idle":"2024-05-14T09:13:25.680544Z","shell.execute_reply.started":"2024-05-14T09:13:22.087719Z","shell.execute_reply":"2024-05-14T09:13:25.679631Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"dataPath = \"/kaggle/input/home-credit-credit-risk-model-stability/\"\n\n# Functions to set dtypes and convert strings\ndef set_table_dtypes(df: pl.DataFrame) -> pl.DataFrame:\n    for col in df.columns:\n        if col[-1] in (\"P\", \"A\"):\n            df = df.with_columns(pl.col(col).cast(pl.Float64).alias(col))\n    return df\n\ndef convert_strings(df: pd.DataFrame) -> pd.DataFrame:\n    for col in df.columns:\n        if df[col].dtype.name in ['object', 'string']:\n            df[col] = df[col].astype(\"string\").astype('category')\n            current_categories = df[col].cat.categories\n            new_categories = current_categories.to_list() + [\"Unknown\"]\n            new_dtype = pd.CategoricalDtype(categories=new_categories, ordered=True)\n            df[col] = df[col].astype(new_dtype)\n    return df\n\n# Load training data\ntrain_basetable = pl.read_csv(dataPath + \"csv_files/train/train_base.csv\")\ntrain_static = pl.concat([\n    pl.read_csv(dataPath + \"csv_files/train/train_static_0_0.csv\").pipe(set_table_dtypes),\n    pl.read_csv(dataPath + \"csv_files/train/train_static_0_1.csv\").pipe(set_table_dtypes),\n], how=\"vertical_relaxed\")\ntrain_static_cb = pl.read_csv(dataPath + \"csv_files/train/train_static_cb_0.csv\").pipe(set_table_dtypes)\ntrain_person_1 = pl.read_csv(dataPath + \"csv_files/train/train_person_1.csv\").pipe(set_table_dtypes)\ntrain_credit_bureau_b_2 = pl.read_csv(dataPath + \"csv_files/train/train_credit_bureau_b_2.csv\").pipe(set_table_dtypes)\n","metadata":{"execution":{"iopub.status.busy":"2024-05-14T09:13:25.682723Z","iopub.execute_input":"2024-05-14T09:13:25.683328Z","iopub.status.idle":"2024-05-14T09:13:39.805632Z","shell.execute_reply.started":"2024-05-14T09:13:25.683291Z","shell.execute_reply":"2024-05-14T09:13:39.804791Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Load test data\ntest_basetable = pl.read_csv(dataPath + \"csv_files/test/test_base.csv\")\ntest_static = pl.concat([\n    pl.read_csv(dataPath + \"csv_files/test/test_static_0_0.csv\").pipe(set_table_dtypes),\n    pl.read_csv(dataPath + \"csv_files/test/test_static_0_1.csv\").pipe(set_table_dtypes),\n    pl.read_csv(dataPath + \"csv_files/test/test_static_0_2.csv\").pipe(set_table_dtypes),\n], how=\"vertical_relaxed\")\ntest_static_cb = pl.read_csv(dataPath + \"csv_files/test/test_static_cb_0.csv\").pipe(set_table_dtypes)\ntest_person_1 = pl.read_csv(dataPath + \"csv_files/test/test_person_1.csv\").pipe(set_table_dtypes)\ntest_credit_bureau_b_2 = pl.read_csv(dataPath + \"csv_files/test/test_credit_bureau_b_2.csv\").pipe(set_table_dtypes)\n\n# Feature engineering\ntrain_person_1_feats_1 = train_person_1.group_by(\"case_id\").agg(\n    pl.col(\"mainoccupationinc_384A\").max().alias(\"mainoccupationinc_384A_max\"),\n    (pl.col(\"incometype_1044T\") == \"SELFEMPLOYED\").max().alias(\"mainoccupationinc_384A_any_selfemployed\")\n)\n\ntrain_person_1_feats_2 = train_person_1.select([\"case_id\", \"num_group1\", \"housetype_905L\"]).filter(\n    pl.col(\"num_group1\") == 0\n).drop(\"num_group1\").rename({\"housetype_905L\": \"person_housetype\"})\n\ntrain_credit_bureau_b_2_feats = train_credit_bureau_b_2.group_by(\"case_id\").agg(\n    pl.col(\"pmts_pmtsoverdue_635A\").max().alias(\"pmts_pmtsoverdue_635A_max\"),\n    (pl.col(\"pmts_dpdvalue_108P\") > 31).max().alias(\"pmts_dpdvalue_108P_over31\")\n)\n\nselected_static_cols = [col for col in train_static.columns if col[-1] in (\"A\", \"M\")]\nselected_static_cb_cols = [col for col in train_static_cb.columns if col[-1] in (\"A\", \"M\")]\n\ndata = train_basetable.join(\n    train_static.select([\"case_id\"] + selected_static_cols), how=\"left\", on=\"case_id\"\n).join(\n    train_static_cb.select([\"case_id\"] + selected_static_cb_cols), how=\"left\", on=\"case_id\"\n).join(\n    train_person_1_feats_1, how=\"left\", on=\"case_id\"\n).join(\n    train_person_1_feats_2, how=\"left\", on=\"case_id\"\n).join(\n    train_credit_bureau_b_2_feats, how=\"left\", on=\"case_id\"\n)\n\n","metadata":{"execution":{"iopub.status.busy":"2024-05-14T09:13:39.807321Z","iopub.execute_input":"2024-05-14T09:13:39.807824Z","iopub.status.idle":"2024-05-14T09:13:41.826106Z","shell.execute_reply.started":"2024-05-14T09:13:39.807794Z","shell.execute_reply":"2024-05-14T09:13:41.825236Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Prepare test features\ntest_person_1_feats_1 = test_person_1.group_by(\"case_id\").agg(\n    pl.col(\"mainoccupationinc_384A\").max().alias(\"mainoccupationinc_384A_max\"),\n    (pl.col(\"incometype_1044T\") == \"SELFEMPLOYED\").max().alias(\"mainoccupationinc_384A_any_selfemployed\")\n)\n\ntest_person_1_feats_2 = test_person_1.select([\"case_id\", \"num_group1\", \"housetype_905L\"]).filter(\n    pl.col(\"num_group1\") == 0\n).drop(\"num_group1\").rename({\"housetype_905L\": \"person_housetype\"})\n\ntest_credit_bureau_b_2_feats = test_credit_bureau_b_2.group_by(\"case_id\").agg(\n    pl.col(\"pmts_pmtsoverdue_635A\").max().alias(\"pmts_pmtsoverdue_635A_max\"),\n    (pl.col(\"pmts_dpdvalue_108P\") > 31).max().alias(\"pmts_dpdvalue_108P_over31\")\n)\n\ndata_submission = test_basetable.join(\n    test_static.select([\"case_id\"] + selected_static_cols), how=\"left\", on=\"case_id\"\n).join(\n    test_static_cb.select([\"case_id\"] + selected_static_cb_cols), how=\"left\", on=\"case_id\"\n).join(\n    test_person_1_feats_1, how=\"left\", on=\"case_id\"\n).join(\n    test_person_1_feats_2, how=\"left\", on=\"case_id\"\n).join(\n    test_credit_bureau_b_2_feats, how=\"left\", on=\"case_id\"\n)\n\n\n\n","metadata":{"execution":{"iopub.status.busy":"2024-05-14T09:13:41.827815Z","iopub.execute_input":"2024-05-14T09:13:41.8281Z","iopub.status.idle":"2024-05-14T09:13:41.840497Z","shell.execute_reply.started":"2024-05-14T09:13:41.828077Z","shell.execute_reply":"2024-05-14T09:13:41.839612Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Check for nulls\ndef total_nulls(df):\n    null_counts = df.select([pl.col(column).is_null().sum().alias(column) for column in df.columns])\n    total_nulls = null_counts.sum(axis=1)[0]\n    return total_nulls\n\nprint(\"Training Data Size:\", train_basetable.shape)\nprint(\"Null Values in Training Data:\", total_nulls(train_basetable))\nprint(\"Testing Data Size:\", data_submission.shape)\nprint(\"Null Values in Testing Data:\", total_nulls(data_submission))\n","metadata":{"execution":{"iopub.status.busy":"2024-05-14T09:13:41.841947Z","iopub.execute_input":"2024-05-14T09:13:41.842274Z","iopub.status.idle":"2024-05-14T09:13:41.865515Z","shell.execute_reply.started":"2024-05-14T09:13:41.842248Z","shell.execute_reply":"2024-05-14T09:13:41.864358Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Split data\ncase_ids = data[\"case_id\"].unique().shuffle(seed=1)\ncase_ids_train, case_ids_test = train_test_split(case_ids, train_size=0.6, random_state=1)\ncase_ids_valid, case_ids_test = train_test_split(case_ids_test, train_size=0.5, random_state=1)\n\ncols_pred = [col for col in data.columns if col[-1].isupper() and col[:-1].islower()]\n\ndef from_polars_to_pandas(case_ids: pl.DataFrame) -> pl.DataFrame:\n    return (\n        data.filter(pl.col(\"case_id\").is_in(case_ids))[[\"case_id\", \"WEEK_NUM\", \"target\"]].to_pandas(),\n        data.filter(pl.col(\"case_id\").is_in(case_ids))[cols_pred].to_pandas(),\n        data.filter(pl.col(\"case_id\").is_in(case_ids))[\"target\"].to_pandas()\n    )\n\nbase_train, X_train, y_train = from_polars_to_pandas(case_ids_train)\nbase_valid, X_valid, y_valid = from_polars_to_pandas(case_ids_valid)\nbase_test, X_test, y_test = from_polars_to_pandas(case_ids_test)\n\nfor df in [X_train, X_valid, X_test]:\n    df = convert_strings(df)\n\nprint(f\"Train: {X_train.shape}\")\nprint(f\"Valid: {X_valid.shape}\")\nprint(f\"Test: {X_test.shape}\")\n\n# One-Hot Encoding and Standard Scaling\nX_train_encoded = pd.get_dummies(X_train, drop_first=True)\nX_valid_encoded = pd.get_dummies(X_valid, drop_first=True)\n\n# Align train and valid set\nX_train_encoded, X_valid_encoded = X_train_encoded.align(X_valid_encoded, join='left', axis=1, fill_value=0)\n\n# Apply StandardScaler\nscaler = StandardScaler()\nX_train_scaled = scaler.fit_transform(X_train_encoded)\nX_valid_scaled = scaler.transform(X_valid_encoded)\n","metadata":{"execution":{"iopub.status.busy":"2024-05-14T09:13:41.866971Z","iopub.execute_input":"2024-05-14T09:13:41.867472Z","iopub.status.idle":"2024-05-14T09:14:57.154169Z","shell.execute_reply.started":"2024-05-14T09:13:41.867438Z","shell.execute_reply":"2024-05-14T09:14:57.15327Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Custom CPPLS objective function\nclass CPPLSObjective:\n    def __init__(self, initial_m=2, variance=1.0, m_steps=100, lr=0.05):\n        self.m = initial_m\n        self.variance = variance\n        self.m_steps = m_steps\n        self.lr = lr\n\n    def update_m(self, y_true, y_pred):\n        m_values = np.linspace(1, 5, self.m_steps)\n        cppls_losses = np.array([self.cppls_loss(y_true, y_pred, m) for m in m_values])\n        cppls_losses -= np.min(cppls_losses)\n        cppls_losses = np.clip(cppls_losses, -100, 100)\n        likelihoods = np.exp(-cppls_losses)\n        priors = norm.pdf(m_values, loc=self.m, scale=np.sqrt(self.variance))\n        posterior = likelihoods * priors\n        posterior_sum = np.sum(posterior)\n        if posterior_sum == 0:\n            posterior += 1e-8\n        posterior /= np.sum(posterior)\n        if np.any(np.isnan(posterior)):\n            posterior = np.nan_to_num(posterior, nan=1.0 / len(posterior))\n        new_m = np.random.choice(m_values, p=posterior)\n        self.m = self.m + self.lr * (new_m - self.m)\n\n    def cppls_loss(self, y_true, y_pred, m):\n        residuals = y_true - y_pred\n        residuals += 1e-8\n        powered_residuals = np.power(np.abs(residuals), m)\n        loss = np.mean(powered_residuals) ** (1 / m)\n        return loss\n\n    def __call__(self, y_true, y_pred):\n        if isinstance(y_true, lgb.Dataset):\n            y_true = y_true.get_label()\n        if isinstance(y_pred, lgb.Dataset):\n            y_pred = y_pred.get_data()\n            if y_pred.ndim > 1:\n                y_pred = y_pred[:, 0]\n        self.update_m(y_true, y_pred)\n        residuals = y_true - y_pred\n        residuals += 1e-8\n        safe_residuals_grad = np.where(residuals != 0, residuals ** (self.m - 1), 0)\n        safe_residuals_hess = np.where(residuals != 0, residuals ** (self.m - 2), 0)\n        gradient = -self.m * safe_residuals_grad / len(y_true)\n        hessian = self.m * (self.m - 1) * safe_residuals_hess / len(y_true)\n        return gradient, hessian\n\n# Prepare the datasets\nlgb_train = lgb.Dataset(X_train_scaled, label=y_train, free_raw_data=False)\nlgb_valid = lgb.Dataset(X_valid_scaled, label=y_valid, reference=lgb_train, free_raw_data=False)\n\n# Parameters\nparams = {\n    \"boosting_type\": \"gbdt\",\n    \"objective\": CPPLSObjective(initial_m=2),\n    \"metric\": \"auc\",\n    \"max_depth\": 3,\n    \"num_leaves\": 31,\n    \"learning_rate\": 0.05,\n    \"feature_fraction\": 0.9,\n    \"bagging_fraction\": 0.8,\n    \"bagging_freq\": 5,\n    \"verbose\": -1,\n}\n\n# Initialize and use the custom objective\ncppls_obj = CPPLSObjective(initial_m=2)\ngbm = lgb.train(\n    params,\n    lgb_train,\n    num_boost_round=1000,\n    valid_sets=[lgb_valid],\n    callbacks=[lgb.log_evaluation(50), lgb.early_stopping(10)]\n)\n\n","metadata":{"execution":{"iopub.status.busy":"2024-05-14T09:14:57.155581Z","iopub.execute_input":"2024-05-14T09:14:57.155851Z","iopub.status.idle":"2024-05-14T09:15:40.302725Z","shell.execute_reply.started":"2024-05-14T09:14:57.155829Z","shell.execute_reply":"2024-05-14T09:15:40.301276Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Evaluation with AUC and stability metrics\nfor base, X in [(base_train, X_train), (base_valid, X_valid), (base_test, X_test)]:\n    y_pred = gbm.predict(X, num_iteration=gbm.best_iteration)\n    base[\"score\"] = y_pred\n\nprint(f'The AUC score on the train set is: {roc_auc_score(base_train[\"target\"], base_train[\"score\"])}')\nprint(f'The AUC score on the valid set is: {roc_auc_score(base_valid[\"target\"], base_valid[\"score\"])}')\nprint(f'The AUC score on the test set is: {roc_auc_score(base_test[\"target\"], base_test[\"score\"])}')\n\ndef gini_stability(base, w_fallingrate=88.0, w_resstd=-0.5):\n    gini_in_time = base.loc[:, [\"WEEK_NUM\", \"target\", \"score\"]]\\\n        .sort_values(\"WEEK_NUM\")\\\n        .groupby(\"WEEK_NUM\")[[\"target\", \"score\"]]\\\n        .apply(lambda x: 2 * roc_auc_score(x[\"target\"], x[\"score\"]) - 1).tolist()\n    x = np.arange(len(gini_in_time))\n    y = gini_in_time\n    a, b = np.polyfit(x, y, 1)\n    y_hat = a * x + b\n    residuals = y - y_hat\n    res_std = np.std(residuals)\n    avg_gini = np.mean(gini_in_time)\n    return avg_gini + w_fallingrate * min(0, a) + w_resstd * res_std\n\nstability_score_train = gini_stability(base_train)\nstability_score_valid = gini_stability(base_valid)\nstability_score_test = gini_stability(base_test)\n\nprint(f'The stability score on the train set is: {stability_score_train}')\nprint(f'The stability score on the valid set is: {stability_score_valid}')\nprint(f'The stability score on the test set is: {stability_score_test}')\n","metadata":{"execution":{"iopub.status.busy":"2024-05-14T09:15:40.303939Z","iopub.execute_input":"2024-05-14T09:15:40.304249Z","iopub.status.idle":"2024-05-14T09:15:41.738193Z","shell.execute_reply.started":"2024-05-14T09:15:40.304223Z","shell.execute_reply":"2024-05-14T09:15:41.736702Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Submission\nX_submission = data_submission[cols_pred].to_pandas()\nX_submission = convert_strings(X_submission)\ncategorical_cols = X_train.select_dtypes(include=['category']).columns\n\nfor col in categorical_cols:\n    train_categories = set(X_train[col].cat.categories)\n    submission_categories = set(X_submission[col].cat.categories)\n    new_categories = submission_categories - train_categories\n    X_submission.loc[X_submission[col].isin(new_categories), col] = \"Unknown\"\n    new_dtype = pd.CategoricalDtype(categories=train_categories, ordered=True)\n    X_train[col] = X_train[col].astype(new_dtype)\n    X_submission[col] = X_submission[col].astype(new_dtype)\n\ny_submission_pred = gbm.predict(X_submission, num_iteration=gbm.best_iteration)\n\nsubmission = pd.DataFrame({\n    \"case_id\": data_submission[\"case_id\"].to_numpy(),\n    \"score\": y_submission_pred\n}).set_index('case_id')\nsubmission.to_csv(\"./submission.csv\")\n# This Python 3 environment comes with many helpful analytics libraries installed\n# It is defined by the kaggle/python Docker image: https://github.com/kaggle/docker-python\n# For example, here's several helpful packages to load\n\nimport numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n\n# Input data files are available in the read-only \"../input/\" directory\n# For example, running this (by clicking run or pressing Shift+Enter) will list all files under the input directory\n\nimport os\nfor dirname, _, filenames in os.walk('/kaggle/input'):\n    for filename in filenames:\n        print(os.path.join(dirname, filename))\n\n# You can write up to 20GB to the current directory (/kaggle/working/) that gets preserved as output when you create a version using \"Save & Run All\" \n# You can also write temporary files to /kaggle/temp/, but they won't be saved outside of the current session","metadata":{"execution":{"iopub.status.busy":"2024-05-14T09:15:41.739576Z","iopub.status.idle":"2024-05-14T09:15:41.740406Z","shell.execute_reply.started":"2024-05-14T09:15:41.740106Z","shell.execute_reply":"2024-05-14T09:15:41.740129Z"},"trusted":true},"execution_count":null,"outputs":[]}]}