loft-br · Jul 1, 2022 · Jul 1, 2022 · Jul 1, 2022 · Jul 4, 2022 · Jul 5, 2022
diff --git a/.gitignore b/.gitignore
@@ -139,3 +139,5 @@ cython_debug/
 
 .DS_Store
 .vscode/
+
+.idea
diff --git a/custom_tests/debug_warnings.py b/custom_tests/debug_warnings.py
@@ -0,0 +1,125 @@
+# from pycox.datasets import metabric
+from lifelines.datasets import load_dd
+import numpy as np
+import pandas as pd
+from xgbse import XGBSEStackedWeibull
+import matplotlib.pyplot as plt
+from xgbse.converters import convert_to_structured
+from xgbse import XGBSEKaplanTree, XGBSEBootstrapEstimator
+from xgbse.metrics import concordance_index, approx_brier_score, dist_calibration_score
+from sklearn.model_selection import train_test_split
+
+plt.style.use('bmh')
+
+
+# to easily plot confidence intervals
+def plot_ci(mean_, upper_ci_, lower_ci_, i=42, title='Probability of survival $P(T \\geq t)$'):
+    # plotting mean and confidence intervals
+    plt.figure(figsize=(12, 4), dpi=120)
+    plt.plot(mean_.columns, mean_.iloc[i])
+    plt.fill_between(mean_.columns, lower_ci_.iloc[i], upper_ci_.iloc[i], alpha=0.2)
+
+    plt.title(title)
+    plt.xlabel('Time [days]')
+    plt.ylabel('Probability')
+    plt.tight_layout()
+
+
+df = load_dd()
+
+# splitting to X, T, E format
+X = df.drop(['duration', 'observed'], axis=1)
+X = X.astype({c: "category" for c in df.columns if df[c].dtype.name == "object"})
+feature_types = ["c" if X[c].dtype.name in ["object", "category"] else "q" for c in X.columns]
+T = df['duration']
+E = df['observed']
+y = convert_to_structured(T, E)
+ENABLE_CATEGORICAL = True
+
+# splitting between train, and validation
+X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=1 / 3, random_state=0)
+TIME_BINS = np.arange(T.min(), T.max(), int((T.max() - T.min())/10 + 1))
+# TIME_BINS
+# ######################################################################################################################
+bootstrap_estimator = XGBSEBootstrapEstimator(
+    XGBSEStackedWeibull(),
+    n_estimators=3
+)
+
+# fitting the meta estimator
+bootstrap_estimator.fit(X_train, y_train,
+                        time_bins=TIME_BINS,
+                        enable_categorical=ENABLE_CATEGORICAL,
+                        validation_data=(X_test, y_test),
+                        early_stopping_rounds=100,
+                        verbose_eval=100,
+                        feature_types=feature_types)
+
+# predicting
+mean_prob, upper_ci_prob, lower_ci_prob = bootstrap_estimator.predict(
+    X_test,
+    return_ci=True,
+    enable_categorical=ENABLE_CATEGORICAL,
+    feature_types=feature_types
+)
+
+print(f"C-index XGBSEStackedWeibull bootstrap: {concordance_index(y_test, mean_prob)}")
+print(f"Avg. Brier Score XGBSEStackedWeibull bootstrap: {approx_brier_score(y_test, mean_prob)}")
+
+d_calib_weibull = dist_calibration_score(y_test, mean_prob, returns='all')
+print(f"D-Calibration XGBSEStackedWeibull: {d_calib_weibull}")
+
+# ######################################################################################################################
+
+
+# xgboost parameters to fit our model
+PARAMS_TREE = {
+    'objective': 'survival:cox',
+    'eval_metric': 'cox-nloglik',
+    'tree_method': 'hist',
+    'max_depth': 10,
+    'booster': 'dart',
+    'subsample': 1.0,
+    'min_child_weight': 50,
+    'colsample_bynode': 1.0
+}
+
+# fitting xgbse model
+xgbse_model = XGBSEKaplanTree(PARAMS_TREE)
+xgbse_model.fit(X_train, y_train, time_bins=TIME_BINS, enable_categorical=ENABLE_CATEGORICAL)
+
+# predicting
+mean, upper_ci, lower_ci = xgbse_model.predict(X_test, return_ci=True, enable_categorical=ENABLE_CATEGORICAL)
+
+# print metrics
+print(f"C-index: {concordance_index(y_test, mean)}")
+print(f"Avg. Brier Score: {approx_brier_score(y_test, mean)}")
+
+# plotting CIs
+plot_ci(mean, upper_ci, lower_ci)
+
+#
+# %%time
+# ######################################################################################################################
+# base model as XGBSEKaplanTree
+base_model = XGBSEKaplanTree(PARAMS_TREE)
+
+# bootstrap meta estimator
+bootstrap_estimator = XGBSEBootstrapEstimator(base_model, n_estimators=100)
+
+# fitting the meta estimator
+bootstrap_estimator.fit(X_train, y_train, time_bins=TIME_BINS, enable_categorical=ENABLE_CATEGORICAL)
+
+# predicting
+mean, upper_ci, lower_ci = bootstrap_estimator.predict(X_test, return_ci=True, enable_categorical=ENABLE_CATEGORICAL)
+
+# print metrics
+print(f"C-index: {concordance_index(y_test, mean)}")
+print(f"Avg. Brier Score: {approx_brier_score(y_test, mean)}")
+
+# plotting CIs
+plot_ci(mean, upper_ci, lower_ci)
+
+# ######################################################################################################################
+# ######################################################################################################################
+print("End of script!")