Fix DataFrame feature name warnings in sklearn wrapper models (#540)

pyod/models/gmm.py

@@ -219,6 +219,7 @@ def decision_function(self, X):
             The anomaly score of the input samples.
         """
         check_is_fitted(self, ["decision_scores_", "threshold_", "labels_"])
+        X = check_array(X)
 
         # Invert outlier scores. Outliers come with higher outlier scores
         return invert_order(self.detector_.score_samples(X))

pyod/models/iforest.py

@@ -241,6 +241,7 @@ def decision_function(self, X):
             The anomaly score of the input samples.
         """
         check_is_fitted(self, ['decision_scores_', 'threshold_', 'labels_'])
+        X = check_array(X)
         # invert outlier scores. Outliers comes with higher outlier scores
         return invert_order(self.detector_.decision_function(X))
 

pyod/models/lof.py

@@ -211,6 +211,7 @@ def decision_function(self, X):
 
         check_is_fitted(self, ['decision_scores_', 'threshold_', 'labels_'])
 
+        X = check_array(X, accept_sparse=True)
         # Invert outlier scores. Outliers comes with higher outlier scores
         # noinspection PyProtectedMember
         try:

pyod/models/ocsvm.py

@@ -188,6 +188,7 @@ def decision_function(self, X):
             The anomaly score of the input samples.
         """
         check_is_fitted(self, ['decision_scores_', 'threshold_', 'labels_'])
+        X = check_array(X)
         # Invert outlier scores. Outliers comes with higher outlier scores
         return invert_order(self.detector_.decision_function(X))
 

pyod/test/test_iforest.py

@@ -167,6 +167,29 @@ def test_feature_importances(self):
         feature_importances = self.clf.feature_importances_
         assert (len(feature_importances) == 2)
 
+    def test_dataframe_no_feature_name_warning(self):
+        """Regression test for GitHub issue #540.
+
+        When a pandas DataFrame is passed to fit/predict, no warning about
+        feature names should be raised by the underlying sklearn estimator.
+        """
+        import pytest
+        import warnings
+
+        pd = pytest.importorskip("pandas")
+
+        df_train = pd.DataFrame(self.X_train, columns=['f1', 'f2'])
+        df_test = pd.DataFrame(self.X_test, columns=['f1', 'f2'])
+
+        clf = IForest(contamination=self.contamination, random_state=42)
+        clf.fit(df_train)
+
+        with warnings.catch_warnings():
+            warnings.simplefilter("error", UserWarning)
+            clf.decision_function(df_test)
+            clf.predict(df_test)
+            clf.predict_proba(df_test)
+
     def tearDown(self):
         pass
 

pyod/test/test_lof.py

@@ -10,6 +10,7 @@
 from numpy.testing import assert_array_less
 from numpy.testing import assert_equal
 from numpy.testing import assert_raises
+from scipy.sparse import csr_matrix
 from scipy.stats import rankdata
 from sklearn.base import clone
 from sklearn.metrics import roc_auc_score
@@ -61,6 +62,16 @@ def test_prediction_scores(self):
         # check performance
         assert (roc_auc_score(self.y_test, pred_scores) >= self.roc_floor)
 
+    def test_sparse_prediction_scores(self):
+        sparse_train = csr_matrix(self.X_train)
+        sparse_test = csr_matrix(self.X_test)
+        clf = LOF(contamination=self.contamination)
+        clf.fit(sparse_train)
+
+        pred_scores = clf.decision_function(sparse_test)
+
+        assert_equal(pred_scores.shape[0], self.X_test.shape[0])
+
     def test_prediction_labels(self):
         pred_labels = self.clf.predict(self.X_test)
         assert_equal(pred_labels.shape, self.y_test.shape)