raise error instead of warning when a user has missing data and add c… (#2143)

* raise error instead of warning when a user has missing data and add check for train data in addition to test

* address comments

Author: hawestra

responsibleai/responsibleai/rai_insights/rai_insights.py

@@ -597,14 +597,9 @@ def _validate_rai_insights_input_parameters(
                 "identified as categorical features: "
                 f"{non_categorical_or_time_string_columns}")
 
-        list_of_feature_having_missing_values = []
-        for feature in test.columns.tolist():
-            if np.any(test[feature].isnull()):
-                list_of_feature_having_missing_values.append(feature)
-        if len(list_of_feature_having_missing_values) > 0:
-            warnings.warn(
-                f"Features {list_of_feature_having_missing_values} "
-                "have missing values in test data")
+        # Check if any of the data is missing in test and train data
+        self._validate_data_is_not_missing(test, "test")
+        self._validate_data_is_not_missing(train, "train")
 
         self._validate_feature_metadata(
             feature_metadata, train, task_type, model, target_column)
@@ -717,6 +712,17 @@ def _validate_classes(
                     if_predictions=True
                 )
 
+    def _validate_data_is_not_missing(self, data, data_name):
+        """Validates that data is not missing (ie null)"""
+        list_of_feature_having_missing_values = []
+        for feature in data.columns.tolist():
+            if np.any(data[feature].isnull()):
+                list_of_feature_having_missing_values.append(feature)
+        if len(list_of_feature_having_missing_values) > 0:
+            raise UserConfigValidationException(
+                f"Features {list_of_feature_having_missing_values} "
+                f"have missing values in {data_name} data.")
+
     def _validate_feature_metadata(
             self, feature_metadata, train, task_type, model, target_column):
         """Validates the feature metadata."""

responsibleai/tests/rai_insights/test_rai_insights_validations.py

@@ -48,7 +48,7 @@ def test_validate_unsupported_task_type(self, forecasting_enabled):
                 task_type='regre',
                 forecasting_enabled=forecasting_enabled)
 
-    def test_missing_data_warnings(self):
+    def test_missing_test_data(self):
         train_data = {
             'Column1': [10, 20, 90, 40, 50],
             'Column2': [10, 20, 90, 40, 50],
@@ -57,7 +57,39 @@ def test_missing_data_warnings(self):
         train = pd.DataFrame(train_data)
 
         test_data = {
-            'Column1': [10, 20, np.nan, 40, 50],
+            'Column1': [10, 20, 90, 40, 50],
+            'Column2': [10, 20, 90, 40, 50],
+            'Target': [10, 20, np.nan, 40, 50]
+        }
+        test = pd.DataFrame(test_data)
+
+        X_train = train.drop(columns=['Target'])
+        y_train = train['Target'].values
+        model = create_complex_classification_pipeline(
+            X_train, y_train, ['Column1', 'Column2'], [])
+
+        with pytest.raises(
+            UserConfigValidationException,
+                match="['Column1']") as ucve:
+            RAIInsights(
+                model=model,
+                train=train,
+                test=test,
+                target_column='Target',
+                task_type='classification')
+        assert "Features ['Target'] have missing values in " + \
+            "test data" in str(ucve.value)
+
+    def test_missing_train_data(self):
+        train_data = {
+            'Column1': [10, 20, 90, 40, 50],
+            'Column2': [10, 20, np.nan, 40, 50],
+            'Target': [10, 20, 90, 40, 50]
+        }
+        train = pd.DataFrame(train_data)
+
+        test_data = {
+            'Column1': [10, 20, 90, 40, 50],
             'Column2': [10, 20, 90, 40, 50],
             'Target': [10, 20, 90, 40, 50]
         }
@@ -68,15 +100,17 @@ def test_missing_data_warnings(self):
         model = create_complex_classification_pipeline(
             X_train, y_train, ['Column1', 'Column2'], [])
 
-        with pytest.warns(
-            UserWarning,
-                match="['Column1']"):
+        with pytest.raises(
+            UserConfigValidationException,
+                match="['Column2']") as ucve:
             RAIInsights(
                 model=model,
                 train=train,
                 test=test,
                 target_column='Target',
                 task_type='classification')
+        assert "Features ['Column2'] have missing values in " + \
+            "train data" in str(ucve.value)
 
     def test_validate_test_data_size(self):
         X_train, X_test, y_train, y_test, _, _ = \