rename mixin (#81)

* rename to base class

* move

* fix

* fmt

* mixin

skranger/ensemble/base.py

@@ -1,3 +1,4 @@
+import bisect
 import warnings
 from collections.abc import Iterable
 
@@ -6,7 +7,7 @@
 from sklearn.utils.validation import check_is_fitted
 
 
-class RangerValidationMixin:
+class RangerMixin:
     @property
     def feature_importances_(self):
         try:
@@ -22,6 +23,38 @@ def feature_importances_(self):
                 "importance must be set to something other than 'none'"
             ) from None
 
+    def get_importance_pvalues(self):
+        """Calculate p-values for variable importances.
+
+        Uses the fast method from Janitza et al. (2016).
+        """
+        check_is_fitted(self)
+        if self.importance != "impurity_corrected":
+            raise ValueError(
+                "p-values can only be calculated with importance parameter set to 'impurity_corrected'"
+            )
+
+        vimp = np.array(self.ranger_forest_["variable_importance"])
+        m1 = vimp[vimp < 0]
+        m2 = vimp[vimp == 0]
+
+        if len(m1) == 0:
+            raise ValueError(
+                "No negative importance values found, cannot calculate p-values."
+            )
+        if len(m2) < 1:
+            vimp_dist = np.concatenate((m1, -m1))
+        else:
+            vimp_dist = np.concatenate((m1, -m1, m2))
+
+        vimp_dist.sort()
+        result = []
+        for i in range(len(vimp)):
+            result.append(bisect.bisect_left(vimp_dist, vimp[i]))
+        pval = 1 - np.array(result) / len(vimp_dist)
+        return pval
+
+    # region validation
     def _validate_parameters(self, X, y, sample_weights):
         """Validate ranger parameters and set defaults."""
         self.n_jobs_ = max(
@@ -212,3 +245,5 @@ def _check_inbag(self, sample_weights):
                 raise ValueError("Cannot use class sampling and inbag.")
             if len(self.inbag) != self.n_estimators:
                 raise ValueError("Size of inbag must be equal to n_estimators.")
+
+    # endregion

skranger/ensemble/ranger_forest_classifier.py

@@ -1,6 +1,4 @@
 """Scikit-learn wrapper for ranger classification."""
-import bisect
-
 import numpy as np
 from sklearn.base import BaseEstimator
 from sklearn.base import ClassifierMixin
@@ -10,10 +8,10 @@
 from sklearn.utils.validation import check_is_fitted
 
 from skranger.ensemble import ranger
-from skranger.ensemble.base import RangerValidationMixin
+from skranger.ensemble.base import RangerMixin
 
 
-class RangerForestClassifier(RangerValidationMixin, ClassifierMixin, BaseEstimator):
+class RangerForestClassifier(RangerMixin, ClassifierMixin, BaseEstimator):
     r"""Ranger Random Forest Probability/Classification implementation for sci-kit learn.
 
     Provides a sklearn classifier interface to the Ranger C++ library using Cython.
@@ -318,38 +316,6 @@ def predict_log_proba(self, X):
         proba = self.predict_proba(X)
         return np.log(proba)
 
-    def get_importance_pvalues(self):
-        """Calculate p-values for variable importance.
-
-        Uses the fast method from Janitza et al. (2016).
-        """
-
-        check_is_fitted(self)
-        if self.importance != "impurity_corrected":
-            raise ValueError(
-                "p-values can only be calculated with importance parameter set to 'impurity_corrected'"
-            )
-
-        vimp = np.array(self.ranger_forest_["variable_importance"])
-        m1 = vimp[vimp < 0]
-        m2 = vimp[vimp == 0]
-
-        if len(m1) == 0:
-            raise ValueError(
-                "No negative importance values found, cannot calculate p-values."
-            )
-        if len(m2) < 1:
-            vimp_dist = np.concatenate((m1, -m1))
-        else:
-            vimp_dist = np.concatenate((m1, -m1, m2))
-
-        vimp_dist.sort()
-        result = []
-        for i in range(len(vimp)):
-            result.append(bisect.bisect_left(vimp_dist, vimp[i]))
-        pval = 1 - np.array(result) / len(vimp_dist)
-        return pval
-
     def _more_tags(self):
         return {
             "_xfail_checks": {

skranger/ensemble/ranger_forest_regressor.py

@@ -1,6 +1,4 @@
 """Scikit-learn wrapper for ranger regression."""
-import bisect
-
 import numpy as np
 from sklearn.base import BaseEstimator
 from sklearn.base import RegressorMixin
@@ -9,10 +7,10 @@
 from sklearn.utils.validation import check_is_fitted
 
 from skranger.ensemble import ranger
-from skranger.ensemble.base import RangerValidationMixin
+from skranger.ensemble.base import RangerMixin
 
 
-class RangerForestRegressor(RangerValidationMixin, RegressorMixin, BaseEstimator):
+class RangerForestRegressor(RangerMixin, RegressorMixin, BaseEstimator):
     r"""Ranger Random Forest Regression implementation for sci-kit learn.
 
     Provides a sklearn regressor interface to the Ranger C++ library using Cython. The
@@ -396,38 +394,6 @@ def predict(self, X):
         )
         return np.array(result["predictions"])
 
-    def get_importance_pvalues(self):
-        """Calculate p-values for variable importance.
-
-        Uses the fast method from Janitza et al. (2016).
-        """
-
-        check_is_fitted(self)
-        if self.importance != "impurity_corrected":
-            raise ValueError(
-                "p-values can only be calculated with importance parameter set to 'impurity_corrected'"
-            )
-
-        vimp = np.array(self.ranger_forest_["variable_importance"])
-        m1 = vimp[vimp < 0]
-        m2 = vimp[vimp == 0]
-
-        if len(m1) == 0:
-            raise ValueError(
-                "No negative importance values found, cannot calculate p-values."
-            )
-        if len(m2) < 1:
-            vimp_dist = np.concatenate((m1, -m1))
-        else:
-            vimp_dist = np.concatenate((m1, -m1, m2))
-
-        vimp_dist.sort()
-        result = []
-        for i in range(len(vimp)):
-            result.append(bisect.bisect_left(vimp_dist, vimp[i]))
-        pval = 1 - np.array(result) / len(vimp_dist)
-        return pval
-
     def _more_tags(self):
         return {
             "_xfail_checks": {

skranger/ensemble/ranger_forest_survival.py

@@ -6,10 +6,10 @@
 from sklearn.utils.validation import check_is_fitted
 
 from skranger.ensemble import ranger
-from skranger.ensemble.base import RangerValidationMixin
+from skranger.ensemble.base import RangerMixin
 
 
-class RangerForestSurvival(RangerValidationMixin, BaseEstimator):
+class RangerForestSurvival(RangerMixin, BaseEstimator):
     r"""Ranger Random Forest Survival implementation for sci-kit survival.
 
     Provides a sksurv interface to the Ranger C++ library using Cython. The

tests/conftest.py

@@ -83,6 +83,25 @@ def boston_X_mod(mod):
         )
 
 
+@pytest.fixture
+def lung_X_mod(lung_X, mod):
+    if mod == "none":
+        return _lung_X[["Age_in_years", "Karnofsky_score"]]
+    elif mod == "random":
+        np.random.seed(42)
+        return np.concatenate((lung_X, np.random.uniform(size=(lung_X.shape))), 1)
+    elif mod == "const":
+        np.random.seed(42)
+        return np.concatenate(
+            (
+                lung_X,
+                np.random.uniform(size=(lung_X.shape)),
+                np.zeros(shape=(lung_X.shape)),
+            ),
+            1,
+        )
+
+
 @pytest.fixture
 def iris_y():
     return _iris_y

tests/ensemble/test_ranger_forest_classifier.py

@@ -134,8 +134,8 @@ def test_importance_pvalues(self, iris_X_mod, iris_y, importance, mod):
 
         # Test error for no non-negative importance values
         if mod == "none":
+            rfc.fit(iris_X_mod, iris_y)
             with pytest.raises(ValueError):
-                rfc.fit(iris_X_mod, iris_y)
                 rfc.get_importance_pvalues()
             return
 

tests/ensemble/test_ranger_forest_regressor.py

@@ -113,8 +113,8 @@ def test_importance_pvalues(self, boston_X_mod, boston_y, importance, mod):
         # Test error for no non-negative importance values
 
         if mod == "none":
+            rfc.fit(boston_X_mod, boston_y)
             with pytest.raises(ValueError):
-                rfc.fit(boston_X_mod, boston_y)
                 rfc.get_importance_pvalues()
             return
 

tests/ensemble/test_ranger_forest_survival.py

@@ -105,6 +105,31 @@ def test_importance(
             else:
                 assert rfs.importance_mode_ == 3
 
+    def test_importance_pvalues(self, lung_X_mod, lung_y, importance, mod):
+        rfs = RangerForestSurvival(importance=importance)
+        np.random.seed(42)
+
+        if importance not in ["none", "impurity", "impurity_corrected", "permutation"]:
+            with pytest.raises(ValueError):
+                rfs.fit(lung_X_mod, lung_y)
+            return
+
+        if not importance == "impurity_corrected":
+            rfs.fit(lung_X_mod, lung_y)
+            with pytest.raises(ValueError):
+                rfs.get_importance_pvalues()
+            return
+
+        # Test error for no non-negative importance values
+        if mod == "none":
+            rfs.fit(lung_X_mod, lung_y)
+            with pytest.raises(ValueError):
+                rfs.get_importance_pvalues()
+            return
+
+        rfs.fit(lung_X_mod, lung_y)
+        assert len(rfs.get_importance_pvalues()) == lung_X_mod.shape[1]
+
     def test_mtry(self, lung_X, lung_y, mtry):
         rfs = RangerForestSurvival(mtry=mtry)