Merge pull request #6 from messier16/fix-bugged_unseen_labels

Fix error in CategoryEncoder transform

Author: fferegrino

m16_mlutils/pipeline/CategoryEncoder.py

@@ -48,6 +48,6 @@ def transform(self, X):
                 o = self.one_hot_encoders[c].transform(
                     values.reshape(len(values), 1))
             except (KeyError, ValueError):
-                o = np.zeros((1, len(self.label_encoders[c].classes_)))
+                o = np.zeros((len(X), len(self.label_encoders[c].classes_)))
             one_hots.append(o)
         return np.concatenate(one_hots, axis=1)

setup.py

@@ -13,7 +13,7 @@
 from setuptools.command.install import install
 
 # Package version
-VERSION = "0.4.3"
+VERSION = "0.4.4"
 
 class VerifyVersionCommand(install):
     """Custom command to verify that the git tag matches our version"""

tests/test_CategoryEncoder.py

@@ -86,14 +86,17 @@ def test_previously_unseen(self):
         })
 
         test = pd.DataFrame({
-            'l': ['z']
+            'l': ['z', 'y']
         })
 
         encoder = CategoryEncoder()
 
-        expected = self.array([[0,0,0]])
+        expected = self.array([
+            [0, 0, 0],
+            [0, 0, 0],
+        ])
 
         encoder.fit(train)
         actual = encoder.transform(test)
 
-        self.assertArrayEqual(expected, actual)
+        self.assertArrayEqual(expected, actual)