FIX property for predict_proba in MultiOutput Classifier (scikit-lear…

…n#15490)

doc/whats_new/v0.22.rst

@@ -657,6 +657,10 @@ Changelog
 - |Fix| :class:`multioutput.MultiOutputClassifier` now has attribute
   ``classes_``. :pr:`14629` by :user:`Agamemnon Krasoulis <agamemnonc>`.
 
+- |Fix| :class:`multioutput.MultiOutputClassifier` now has `predict_proba`
+  as property and can be checked with `hasattr`.
+  :issue:`15488` :pr:`15490` by :user:`Rebekah Kim <rebekahkim>`
+
 :mod:`sklearn.naive_bayes`
 ...............................
 

sklearn/multioutput.py

@@ -360,7 +360,8 @@ def fit(self, X, Y, sample_weight=None):
         self.classes_ = [estimator.classes_ for estimator in self.estimators_]
         return self
 
-    def predict_proba(self, X):
+    @property
+    def predict_proba(self):
         """Probability estimates.
         Returns prediction probabilities for each class of each output.
 
@@ -382,9 +383,11 @@ def predict_proba(self, X):
         check_is_fitted(self)
         if not all([hasattr(estimator, "predict_proba")
                     for estimator in self.estimators_]):
-            raise ValueError("The base estimator should implement "
-                             "predict_proba method")
+            raise AttributeError("The base estimator should "
+                                 "implement predict_proba method")
+        return self._predict_proba
 
+    def _predict_proba(self, X):
         results = [estimator.predict_proba(X) for estimator in
                    self.estimators_]
         return results

sklearn/tests/test_multioutput.py

@@ -175,6 +175,22 @@ def test_multi_output_classification_partial_fit_parallelism():
         assert est1 is not est2
 
 
+# check multioutput has predict_proba
+def test_hasattr_multi_output_predict_proba():
+    # default SGDClassifier has loss='hinge'
+    # which does not expose a predict_proba method
+    sgd_linear_clf = SGDClassifier(random_state=1, max_iter=5)
+    multi_target_linear = MultiOutputClassifier(sgd_linear_clf)
+    multi_target_linear.fit(X, y)
+    assert not hasattr(multi_target_linear, "predict_proba")
+
+    # case where predict_proba attribute exists
+    sgd_linear_clf = SGDClassifier(loss='log', random_state=1, max_iter=5)
+    multi_target_linear = MultiOutputClassifier(sgd_linear_clf)
+    multi_target_linear.fit(X, y)
+    assert hasattr(multi_target_linear, "predict_proba")
+
+
 # check predict_proba passes
 def test_multi_output_predict_proba():
     sgd_linear_clf = SGDClassifier(random_state=1, max_iter=5)
@@ -199,7 +215,7 @@ def custom_scorer(estimator, X, y):
     multi_target_linear = MultiOutputClassifier(sgd_linear_clf)
     multi_target_linear.fit(X, y)
     err_msg = "The base estimator should implement predict_proba method"
-    with pytest.raises(ValueError, match=err_msg):
+    with pytest.raises(AttributeError, match=err_msg):
         multi_target_linear.predict_proba(X)
 
 
@@ -378,7 +394,8 @@ def test_multi_output_exceptions():
     # and predict_proba are called
     moc = MultiOutputClassifier(LinearSVC(random_state=0))
     assert_raises(NotFittedError, moc.predict, y)
-    assert_raises(NotFittedError, moc.predict_proba, y)
+    with pytest.raises(NotFittedError):
+        moc.predict_proba
     assert_raises(NotFittedError, moc.score, X, y)
     # ValueError when number of outputs is different
     # for fit and score