Fix automl to treat binary as categorical when missing values present (…

…#1292)

ludwig/automl/base_config.py

@@ -241,7 +241,7 @@ def get_field_metadata(
     metadata = []
     for idx, field in enumerate(fields):
         missing_value_percent = 1 - float(field.nonnull_values) / row_count
-        dtype = infer_type(field)
+        dtype = infer_type(field, missing_value_percent)
         metadata.append(
             FieldMetadata(
                 name=field.name,
@@ -278,19 +278,21 @@ def get_field_metadata(
 
 
 def infer_type(
-    field: FieldInfo
+    field: FieldInfo,
+    missing_value_percent: float,
 ) -> str:
     """
     Perform type inference on field
 
     # Inputs
     :param field: (FieldInfo) object describing field
+    :param missing_value_percent: (float) percent of missing values in the column
 
     # Return
     :return: (str) feature type
     """
     distinct_values = field.distinct_values
-    if distinct_values == 2:
+    if distinct_values == 2 and missing_value_percent == 0:
         return BINARY
 
     if field.image_values >= 3:

tests/ludwig/automl/test_base_config.py

@@ -8,22 +8,23 @@
 TARGET_NAME = 'target'
 
 
-@pytest.mark.parametrize("distinct_values,avg_words,img_values,expected", [
-    (ROW_COUNT, 0, 0, NUMERICAL),
-    (10, 0, 0, CATEGORY),
-    (2, 0, 0, BINARY),
-    (ROW_COUNT, 3, 0, TEXT),
-    (ROW_COUNT, 1, ROW_COUNT, IMAGE),
+@pytest.mark.parametrize("distinct_values,avg_words,img_values,missing_vals,expected", [
+    (ROW_COUNT, 0, 0, 0.0, NUMERICAL),
+    (10, 0, 0, 0.0, CATEGORY),
+    (2, 0, 0, 0.0, BINARY),
+    (2, 0, 0, 0.1, CATEGORY),
+    (ROW_COUNT, 3, 0, 0.0, TEXT),
+    (ROW_COUNT, 1, ROW_COUNT, 0.0, IMAGE),
 ])
-def test_infer_type(distinct_values, avg_words, img_values, expected):
+def test_infer_type(distinct_values, avg_words, img_values, missing_vals, expected):
     field = FieldInfo(
         name='foo',
         dtype='object',
         distinct_values=distinct_values,
         avg_words=avg_words,
         image_values=img_values,
     )
-    assert infer_type(field) == expected
+    assert infer_type(field, missing_vals) == expected
 
 
 @pytest.mark.parametrize("idx,distinct_values,dtype,name,expected", [