Add compatibility for Python 3.12 (#239)

Fixes the deprecation/renaming of SHAP code

.github/workflows/cronjob_unit_tests.yml

@@ -24,7 +24,7 @@ jobs:
           - build: windows
             os: windows-latest
             SKIP_LIGHTGBM: False
-        python-version: [3.8, 3.9, "3.10", "3.11"]
+        python-version: [3.8, 3.9, "3.10", "3.11", "3.12"]
     steps:
     - uses: actions/checkout@master
 

.github/workflows/unit_tests.yml

@@ -23,7 +23,7 @@ jobs:
           - build: windows
             os: windows-latest
             SKIP_LIGHTGBM: False
-        python-version: [3.8, 3.9, "3.10", "3.11"]
+        python-version: [3.8, 3.9, "3.10", "3.11", "3.12"]
     steps:
     - uses: actions/checkout@master
 

.pre-commit-config.yaml

@@ -35,7 +35,7 @@ repos:
     hooks:
     - id: ruff-check
       name: 'Ruff: Check for errors, styling issues and complexity, and fixes issues if possible (including import order)'
-      entry: ruff
+      entry: ruff check
       language: system
       args: [ --fix, --no-cache ]
     - id: ruff-format

probatus/utils/shap_helpers.py

@@ -23,7 +23,7 @@
 import numpy as np
 import pandas as pd
 from shap import Explainer
-from shap.explainers._tree import Tree
+from shap.explainers import TreeExplainer
 from shap.utils import sample
 from sklearn.pipeline import Pipeline
 
@@ -59,10 +59,10 @@ def shap_calc(
             - 51 - 100 - shows other warnings and prints
             - above 100 - presents all prints and all warnings (including SHAP warnings).
 
-         approximate (boolean):
+        approximate (boolean):
             if True uses shap approximations - less accurate, but very fast. It applies to tree-based explainers only.
 
-         check_additivity (boolean):
+        check_additivity (boolean):
             if False SHAP will disable the additivity check for tree-based models.
 
         **shap_kwargs: kwargs of the shap.Explainer
@@ -104,9 +104,13 @@ def shap_calc(
             explainer = Explainer(model, masker=mask, **shap_kwargs)
 
         # For tree-explainers allow for using check_additivity and approximate arguments
-        if isinstance(explainer, Tree):
-            # Calculate Shap values
+        if isinstance(explainer, TreeExplainer):
             shap_values = explainer.shap_values(X, check_additivity=check_additivity, approximate=approximate)
+
+            # From SHAP version 0.43+ https://github.com/shap/shap/pull/3121 required to
+            # get the second dimension of calculated Shap values.
+            if not isinstance(shap_values, list) and len(shap_values.shape) == 3:
+                shap_values = shap_values[:, :, 1]
         else:
             # Calculate Shap values
             shap_values = explainer.shap_values(X)

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "probatus"
-version = "3.0.0"
+version = "3.0.1"
 requires-python= ">=3.8"
 description = "Validation of binary classifiers and data used to develop them"
 readme = { file = "README.md", content-type = "text/markdown" }
@@ -20,6 +20,7 @@ classifiers = [
     "Programming Language :: Python :: 3.9",
     "Programming Language :: Python :: 3.10",
     "Programming Language :: Python :: 3.11",
+    "Programming Language :: Python :: 3.12",
     "Topic :: Scientific/Engineering",
     "Topic :: Scientific/Engineering :: Artificial Intelligence",
     "License :: OSI Approved :: MIT License",
@@ -32,7 +33,8 @@ dependencies = [
     "scipy>=1.4.0",
     "joblib>=0.13.2",
     "tqdm>=4.41.0",
-    "shap>=0.41.0,<0.43.0",
+    "shap==0.43.0 ; python_version == '3.8'",
+    "shap>=0.43.0 ; python_version != '3.8'",
     "numpy>=1.23.2",
     "numba>=0.57.0",
 ]

tests/feature_elimination/test_feature_elimination.py

@@ -5,7 +5,6 @@
 from sklearn.datasets import make_classification
 from sklearn.ensemble import RandomForestClassifier
 from sklearn.linear_model import LogisticRegression
-from sklearn.metrics import get_scorer
 from sklearn.model_selection import RandomizedSearchCV, StratifiedGroupKFold, StratifiedKFold
 from sklearn.pipeline import Pipeline
 from sklearn.preprocessing import StandardScaler
@@ -314,7 +313,7 @@ def test_get_feature_shap_values_per_fold(X, y):
     Test with ShapRFECV with features per fold.
     """
     clf = DecisionTreeClassifier(max_depth=1)
-    shap_elimination = ShapRFECV(clf)
+    shap_elimination = ShapRFECV(clf, scoring="roc_auc")
     (
         shap_values,
         train_score,
@@ -325,7 +324,6 @@ def test_get_feature_shap_values_per_fold(X, y):
         clf,
         train_index=[2, 3, 4, 5, 6, 7],
         val_index=[0, 1],
-        scorer=get_scorer("roc_auc"),
     )
     assert test_score == 1
     assert train_score > 0.9
@@ -545,7 +543,7 @@ def test_get_feature_shap_values_per_fold_early_stopping_lightGBM(complex_data):
     X, y = complex_data
     y = preprocess_labels(y, y_name="y", index=X.index)
 
-    shap_elimination = EarlyStoppingShapRFECV(clf, early_stopping_rounds=5)
+    shap_elimination = EarlyStoppingShapRFECV(clf, early_stopping_rounds=5, scoring="roc_auc")
     (
         shap_values,
         train_score,
@@ -556,7 +554,6 @@ def test_get_feature_shap_values_per_fold_early_stopping_lightGBM(complex_data):
         clf,
         train_index=list(range(5, 50)),
         val_index=[0, 1, 2, 3, 4],
-        scorer=get_scorer("roc_auc"),
     )
     assert test_score > 0.6
     assert train_score > 0.6
@@ -573,7 +570,7 @@ def test_get_feature_shap_values_per_fold_early_stopping_CatBoost(complex_data,
     X["f1_categorical"] = X["f1_categorical"].astype(str).astype("category")
     y = preprocess_labels(y, y_name="y", index=X.index)
 
-    shap_elimination = EarlyStoppingShapRFECV(clf, early_stopping_rounds=5)
+    shap_elimination = EarlyStoppingShapRFECV(clf, early_stopping_rounds=5, scoring="roc_auc")
     (
         shap_values,
         train_score,
@@ -584,7 +581,6 @@ def test_get_feature_shap_values_per_fold_early_stopping_CatBoost(complex_data,
         clf,
         train_index=list(range(5, 50)),
         val_index=[0, 1, 2, 3, 4],
-        scorer=get_scorer("roc_auc"),
     )
     assert test_score > 0
     assert train_score > 0.6
@@ -603,7 +599,7 @@ def test_get_feature_shap_values_per_fold_early_stopping_XGBoost(complex_data):
     X["f1_categorical"] = X["f1_categorical"].astype(float)
     y = preprocess_labels(y, y_name="y", index=X.index)
 
-    shap_elimination = EarlyStoppingShapRFECV(clf, early_stopping_rounds=5)
+    shap_elimination = EarlyStoppingShapRFECV(clf, early_stopping_rounds=5, scoring="roc_auc")
     (
         shap_values,
         train_score,
@@ -614,7 +610,6 @@ def test_get_feature_shap_values_per_fold_early_stopping_XGBoost(complex_data):
         clf,
         train_index=list(range(5, 50)),
         val_index=[0, 1, 2, 3, 4],
-        scorer=get_scorer("roc_auc"),
     )
     assert test_score > 0
     assert train_score > 0.6