FIX Fixes duck typing in VotingClassifier (#14287)

thomasjpfan · qinhanmin2014 · commit b55d12e60369 · 2019-07-12T22:03:57.000+08:00
diff --git a/doc/whats_new/v0.22.rst b/doc/whats_new/v0.22.rst
@@ -69,15 +69,18 @@ Changelog
 :mod:`sklearn.ensemble`
 .......................
 
+- |Feature| :class:`ensemble.HistGradientBoostingClassifier` and
+  :class:`ensemble.HistGradientBoostingRegressor` have an additional
+  parameter called `warm_start` that enables warm starting. :pr:`14012` by
+  :user:`Johann Faouzi <johannfaouzi>`.
+
 - |Fix| :class:`ensemble.HistGradientBoostingClassifier` and
   :class:`ensemble.HistGradientBoostingRegressor` now bin the training and
   validation data separately to avoid any data leak. :pr:`13933` by
   `Nicolas Hug`_.
 
-- |Feature| :class:`ensemble.HistGradientBoostingClassifier` and
-  :class:`ensemble.HistGradientBoostingRegressor` have an additional
-  parameter called `warm_start` that enables warm starting. :pr:`14012` by
-  :user:`Johann Faouzi <johannfaouzi>`.
+- |Fix| :func:`ensemble.VotingClassifier.predict_proba` will no longer be
+  present when `voting='hard'`. :pr:`14287` by `Thomas Fan`_.
 
 - |Enhancement| :class:`ensemble.HistGradientBoostingClassifier` the training
   loss or score is now monitored on a class-wise stratified subsample to
diff --git a/sklearn/ensemble/tests/test_voting.py b/sklearn/ensemble/tests/test_voting.py
@@ -68,7 +68,12 @@ def test_predictproba_hardvoting():
                                         ('lr2', LogisticRegression())],
                             voting='hard')
     msg = "predict_proba is not available when voting='hard'"
-    assert_raise_message(AttributeError, msg, eclf.predict_proba, X)
+    with pytest.raises(AttributeError, match=msg):
+        eclf.predict_proba
+
+    assert not hasattr(eclf, "predict_proba")
+    eclf.fit(X, y)
+    assert not hasattr(eclf, "predict_proba")
 
 
 def test_notfitted():
diff --git a/sklearn/ensemble/voting.py b/sklearn/ensemble/voting.py
@@ -313,9 +313,6 @@ def _collect_probas(self, X):
 
     def _predict_proba(self, X):
         """Predict class probabilities for X in 'soft' voting """
-        if self.voting == 'hard':
-            raise AttributeError("predict_proba is not available when"
-                                 " voting=%r" % self.voting)
         check_is_fitted(self, 'estimators_')
         avg = np.average(self._collect_probas(X), axis=0,
                          weights=self._weights_not_none)
@@ -335,6 +332,9 @@ def predict_proba(self):
         avg : array-like, shape (n_samples, n_classes)
             Weighted average probability for each class per sample.
         """
+        if self.voting == 'hard':
+            raise AttributeError("predict_proba is not available when"
+                                 " voting=%r" % self.voting)
         return self._predict_proba
 
     def transform(self, X):
