add fix for sklearn 0.19 _BasePipeline class no longer exists

xcessiv/exceptions.py

@@ -12,3 +12,6 @@ def to_dict(self):
         rv = dict(self.kwargs or ())
         rv['message'] = self.message
         return rv
+
+    def __repr__(self):
+        return self.message

xcessiv/stacker.py

@@ -1,8 +1,12 @@
-from sklearn.pipeline import _BasePipeline
+import sklearn
+if sklearn.__version__.startswith('0.18'):
+    from sklearn.pipeline import _BasePipeline as bp
+else:
+    from sklearn.utils.metaestimators import _BaseComposition as bp
 import numpy as np
 
 
-class XcessivStackedEnsemble(_BasePipeline):
+class XcessivStackedEnsemble(bp):
     """Contains the class for the Xcessiv stacked ensemble"""
     def __init__(self, base_learners, meta_feature_generators,
                  secondary_learner, cv_function):

xcessiv/tests/test_functions.py

@@ -189,14 +189,15 @@ def test_verify_estimator_class(self):
             'min_weight_fraction_leaf': 0.0,
             'criterion': 'gini',
             'random_state': None,
-            'min_impurity_split': 1e-07,
+            'min_impurity_split': None,
+            'min_impurity_decrease': 0.0,
             'max_features': 'auto',
             'max_depth': None,
             'class_weight': None
         }
 
     def test_non_serializable_parameters(self):
-        pipeline = Pipeline((('pca', PCA()), ('rf', RandomForestClassifier())))
+        pipeline = Pipeline([('pca', PCA()), ('rf', RandomForestClassifier())])
         performance_dict, hyperparameters = functions.verify_estimator_class(
             pipeline,
             'predict_proba',

xcessiv/tests/test_stacker.py

@@ -0,0 +1,52 @@
+from __future__ import absolute_import, print_function, division, unicode_literals
+import unittest
+import numpy as np
+from xcessiv import stacker
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.linear_model import LogisticRegression
+from sklearn.model_selection import StratifiedKFold, train_test_split
+from sklearn.datasets import load_iris
+from sklearn.metrics import accuracy_score
+
+
+class TestStacker(unittest.TestCase):
+    def setUp(self):
+        bl1 = RandomForestClassifier(random_state=8)
+        bl2 = LogisticRegression()
+        bl3 = RandomForestClassifier(max_depth=10, random_state=10)
+
+        meta_est = LogisticRegression()
+
+        skf = StratifiedKFold(random_state=8).split
+
+        self.stacked_ensemble = stacker.XcessivStackedEnsemble(
+            [bl1, bl2, bl3],
+            ['predict', 'predict_proba', 'predict_proba'],
+            meta_est,
+            skf
+        )
+
+    def test_fit_and_process_using_meta_feature_generator(self):
+        X, y = load_iris(return_X_y=True)
+        X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=8)
+
+        self.stacked_ensemble.fit(X_train, y_train)
+
+        preds = self.stacked_ensemble._process_using_meta_feature_generator(X_test, 'predict')
+        assert np.round(accuracy_score(y_test, preds), 3) == 0.868
+
+        probas = self.stacked_ensemble._process_using_meta_feature_generator(X_test, 'predict_proba')
+        preds = np.argmax(probas, axis=1)
+        assert np.round(accuracy_score(y_test, preds), 3) == 0.868
+
+    def test_get_params(self):
+        self.stacked_ensemble.get_params()
+
+    def test_set_params(self):
+        self.stacked_ensemble.set_params(bl0__random_state=20)
+        assert self.stacked_ensemble.get_params()['bl0__random_state'] == 20
+        assert self.stacked_ensemble.get_params()['bl0'].get_params()['random_state'] == 20
+
+        self.stacked_ensemble.set_params(**{'secondary-learner__C': 1.5})
+        assert self.stacked_ensemble.get_params()['secondary-learner__C'] == 1.5
+        assert self.stacked_ensemble.get_params()['secondary-learner'].get_params()['C'] == 1.5