resolve adding shap values to df

TorkamaniLab · Jan 29, 2024 · 4b6e5f2 · 4b6e5f2
1 parent 84fc151
commit 4b6e5f2
Show file tree

Hide file tree

Showing 6 changed files with 2,261 additions and 150 deletions.
diff --git a/examples/shap/shap_estimator_optimized_by_gridsearchcv.ipynb b/examples/shap/shap_estimator_optimized_by_gridsearchcv.ipynb
diff --git a/examples/shap/shap_estimator_optimized_by_gridsearchcv_fastreeshap.ipynb b/examples/shap/shap_estimator_optimized_by_gridsearchcv_fastreeshap.ipynb
diff --git a/examples/shap/shap_estimator_optimized_by_gridsearchcv_gpboost.ipynb b/examples/shap/shap_estimator_optimized_by_gridsearchcv_gpboost.ipynb
diff --git a/tests/test_gpboost.py b/tests/test_gpboost.py
@@ -1,14 +1,26 @@
 import pytest
 import pandas as pd
 import numpy as np
+import sys
+import subprocess
+
 from sklearn.datasets import make_classification, make_regression
 from sklearn.model_selection import KFold
 import time
 from zoish.feature_selectors.shap_selectors import ShapFeatureSelector
-import gpboost as gpb
+#import gpboost as gpb
 from sklearn.metrics import f1_score, r2_score
 
+def import_gpboost():
+    try:
+        import gpboost as gpb
+    except ImportError:
+        subprocess.check_call([sys.executable, '-m', 'pip', 'install', 'gpboost==1.2.6'])
+        import gpboost as gpb
+    return gpb
+
 
+gpb = import_gpboost()
 @pytest.fixture
 def binary_classification_dataset_with_random_effects():
     # Generate classification data
@@ -95,6 +107,7 @@ def test_shap_feature_selector_binary_classification_with_random_effects(model_c
 
 @pytest.mark.parametrize("model_class", [gpb.GPBoostRegressor])
 def test_shap_feature_selector_regression_with_random_effects(model_class, regression_dataset_with_random_effects):
+    gpb = import_gpboost()
     X, y, groups = regression_dataset_with_random_effects
     gp_model = gpb.GPModel(group_data=groups, likelihood="gaussian")
     gp_model.set_prediction_data(group_data_pred=groups)

diff --git a/tests/test_zoish.py b/tests/test_zoish.py
@@ -1,3 +1,3 @@
 from zoish import __version__
 def test_version():
-    assert __version__ == "5.0.2"
+    assert __version__ == "5.0.3"
diff --git a/zoish/feature_selectors/shap_selectors.py b/zoish/feature_selectors/shap_selectors.py
@@ -4,12 +4,8 @@
 import logging
 import warnings
 
-# import sys
-# import subprocess
-
 import fasttreeshap
 
-
 # Plotting libraries
 import numpy as np
 import pandas as pd
@@ -23,15 +19,6 @@
 
 logger.info("Shap Feature Selector has started !")
 
-
-# def import_gpboost():
-#     try:
-#         import gpboost as gpb
-#     except ImportError:
-#         subprocess.check_call([sys.executable, '-m', 'pip', 'install', 'gpboost==1.2.6'])
-#         import gpboost as gpb
-#     return gpb
-
 class ShapPlotFeatures(PlotFeatures):
     """
     Initializes the class with a feature selector and additional keyword arguments.
@@ -696,7 +683,10 @@ def setup_kernel_explainer(X):
 
                     # Implement the functions
                     class_name = self.model.__class__.__name__
-                    is_gpboost_model = "GPBoostClassifier" in class_name or "GPBoostRegressor" in class_name
+                    is_gpboost_model = (
+                        "GPBoostClassifier" in class_name
+                        or "GPBoostRegressor" in class_name
+                    )
                     if not is_gpboost_model:
                         setup_kernel_explainer(X)
                         self.shap_values = self.explainer.shap_values(X)
@@ -713,6 +703,9 @@ def setup_kernel_explainer(X):
                     self.model, **self.shap_fast_tree_explainer_kwargs
                 )
                 self.shap_values = self.explainer.shap_values(X)
+                print(
+                    "FastTreeShap TreeExplainer has used !"
+                )
             except Exception as e:
                 logger.error(f"FastTreeShap TreeExplainer could not be used: {e}")
                 raise e
@@ -765,18 +758,18 @@ def setup_kernel_explainer(X):
             )
         if self.feature_importances_ is not None:
             ordered_importances = self.feature_importances_[self.importance_order]
-            self.importance_df['Values'] = ordered_importances
+            self.importance_df["Values"] = ordered_importances
 
         else:
-            raise ValueError(
-                "feature_importances_ is None."
-            )
+            raise ValueError("feature_importances_ is None.")
 
         self.X = X
         self.y = y
 
         class_name = self.model.__class__.__name__
-        is_gpboost_model = "GPBoostClassifier" in class_name or "GPBoostRegressor" in class_name
+        is_gpboost_model = (
+            "GPBoostClassifier" in class_name or "GPBoostRegressor" in class_name
+        )
         if is_gpboost_model:
             return self
         else: