working with the xgboost libraries.

TODO:
- check results times or probabilities
- integrate in patient-similarity notebook

xgbse/_debiased_bce.py

@@ -234,7 +234,7 @@ def fit(
         )
         self.feature_importances_ = self.bst.get_score()
         # predicting and encoding leaves
-        self.encoder = OneHotEncoder()
+        self.encoder = OneHotEncoder(handle_unknown="ignore")
         leaves = self.bst.predict(
             dtrain, pred_leaf=True, iteration_range=(0, self.bst.best_iteration + 1)
         )

xgbse/_kaplan_neighbors.py

@@ -504,11 +504,11 @@ def predict(self, X, return_ci=False, return_interval_probs=False, enable_catego
         leaves = self.bst.predict(
             d_matrix, pred_leaf=True, iteration_range=(0, self.bst.best_iteration + 1)
         )
-
+        leaves_index = [x for x in leaves if x in self._train_survival.index]
         # searching for kaplan meier curves in leaves
-        preds_df = self._train_survival.loc[leaves].reset_index(drop=True)
-        upper_ci = self._train_upper_ci.loc[leaves].reset_index(drop=True)
-        lower_ci = self._train_lower_ci.loc[leaves].reset_index(drop=True)
+        preds_df = self._train_survival.loc[leaves_index].reset_index(drop=True)
+        upper_ci = self._train_upper_ci.loc[leaves_index].reset_index(drop=True)
+        lower_ci = self._train_lower_ci.loc[leaves_index].reset_index(drop=True)
 
         if return_ci and return_interval_probs:
             raise ValueError(

xgbse/_meta.py

@@ -60,7 +60,7 @@ def fit(self, X, y, **kwargs):
 
         return self
 
-    def predict(self, X, return_ci=False, ci_width=0.683, return_interval_probs=False):
+    def predict(self, X, return_ci=False, ci_width=0.683, return_interval_probs=False, enable_categorical=False):
 
         """
         Predicts survival as given by the base estimator. A survival function, its upper and lower
@@ -73,6 +73,13 @@ def predict(self, X, return_ci=False, ci_width=0.683, return_interval_probs=Fals
 
             ci_width (Float): width of confidence interval
 
+            enable_categorical: boolean, optional
+                .. versionadded:: 1.3.0
+                .. note:: This parameter is experimental
+                Experimental support of specializing for categorical features.  Do not set
+                to True unless you are interested in development. Also, JSON/UBJSON
+                serialization format is required.
+
         Returns:
             ([(pd.DataFrame, np.array, np.array), pd.DataFrame]):
             preds_df: A dataframe of survival probabilities
@@ -92,7 +99,7 @@ def predict(self, X, return_ci=False, ci_width=0.683, return_interval_probs=Fals
         for estimator in self.estimators_:
 
             temp_preds = estimator.predict(
-                X, return_interval_probs=return_interval_probs
+                X, return_interval_probs=return_interval_probs, enable_categorical=enable_categorical
             )
             preds_list.append(temp_preds)
 

xgbse/metrics.py

@@ -147,7 +147,7 @@ def approx_brier_score(y_true, survival, aggregate="mean"):
     # adding censoring distribution survival at event
     event_time_windows = _match_times_to_windows(times, survival.columns)
     scoring_df["cens_at_event"] = censoring_dist[event_time_windows].iloc[0].values
-
+    # TODO Something is broken when use sklearn.model_selection.cross_val_score
     # list of window results
     window_results = []