Prediction: Output error

Author: VesnaT

Orange/widgets/evaluate/owpredictions.py

@@ -265,6 +265,12 @@ def class_var(self):
     def is_discrete_class(self):
         return bool(self.class_var) and self.class_var.is_discrete
 
+    @property
+    def shown_errors(self):
+        return self.class_var and (
+            self.show_probability_errors if self.is_discrete_class
+            else self.show_reg_errors != NO_ERR)
+
     @Inputs.predictors
     def set_predictor(self, index, predictor: Model):
         item = self.predictors[index]
@@ -331,15 +337,14 @@ def _reg_error_changed(self):
         self._update_prediction_delegate()
 
     def _update_errors_visibility(self):
-        shown = self.class_var and (
-             self.show_probability_errors if self.is_discrete_class
-             else self.show_reg_errors != NO_ERR)
+        shown = self.shown_errors
         view = self.predictionsview
         for col, slot in enumerate(self.predictors):
             view.setColumnHidden(
                 2 * col + 1,
                 not shown or
                 self.is_discrete_class is not slot.predictor.domain.has_discrete_class)
+        self._commit_predictions()
 
     def _set_class_values(self):
         self.class_values = []
@@ -814,12 +819,12 @@ def _commit_predictions(self):
 
         newmetas = []
         newcolumns = []
-        for slot in self._non_errored_predictors():
+        for i, slot in enumerate(self._non_errored_predictors()):
             target = slot.predictor.domain.class_var
             if target and target.is_discrete:
-                self._add_classification_out_columns(slot, newmetas, newcolumns)
+                self._add_classification_out_columns(slot, newmetas, newcolumns, i)
             else:
-                self._add_regression_out_columns(slot, newmetas, newcolumns)
+                self._add_regression_out_columns(slot, newmetas, newcolumns, i)
 
         attrs = list(self.data.domain.attributes)
         metas = list(self.data.domain.metas)
@@ -857,7 +862,7 @@ def _commit_predictions(self):
             predictions = predictions[datamodel.mapToSourceRows(...)]
         self.Outputs.predictions.send(predictions)
 
-    def _add_classification_out_columns(self, slot, newmetas, newcolumns):
+    def _add_classification_out_columns(self, slot, newmetas, newcolumns, index):
         pred = slot.predictor
         name = pred.name
         values = pred.domain.class_var.values
@@ -877,10 +882,21 @@ def _add_classification_out_columns(self, slot, newmetas, newcolumns):
             else:
                 newcolumns.append(numpy.zeros(probs.shape[0]))
 
-    @staticmethod
-    def _add_regression_out_columns(slot, newmetas, newcolumns):
+        # Column with error
+        self._add_error_out_columns(slot, newmetas, newcolumns, index)
+
+    def _add_regression_out_columns(self, slot, newmetas, newcolumns, index):
         newmetas.append(ContinuousVariable(name=slot.predictor.name))
         newcolumns.append(slot.results.unmapped_predicted)
+        self._add_error_out_columns(slot, newmetas, newcolumns, index)
+
+    def _add_error_out_columns(self, slot, newmetas, newcolumns, index):
+        if self.shown_errors:
+            name = f"{slot.predictor.name} (error)"
+            newmetas.append(ContinuousVariable(name=name))
+            err = self.predictionsview.model().errorColumn(index)
+            err[err == 2] = numpy.nan
+            newcolumns.append(err)
 
     def send_report(self):
         def merge_data_with_predictions():

Orange/widgets/evaluate/tests/test_owpredictions.py

@@ -869,6 +869,7 @@ def test_output_wrt_shown_probs_1(self):
         self.send_signal(widget.Inputs.predictors, bayes01, 0)
         self.send_signal(widget.Inputs.predictors, bayes12, 1)
         self.send_signal(widget.Inputs.predictors, bayes012, 2)
+        widget.controls.show_probability_errors.setChecked(False)
 
         for i, pred in enumerate(widget.predictors):
             p = pred.results.unmapped_probabilities
@@ -918,6 +919,7 @@ def test_output_wrt_shown_probs_2(self):
         self.send_signal(widget.Inputs.data, iris012)
         self.send_signal(widget.Inputs.predictors, bayes01, 0)
         self.send_signal(widget.Inputs.predictors, bayes012, 1)
+        widget.controls.show_probability_errors.setChecked(False)
 
         for i, pred in enumerate(widget.predictors):
             p = pred.results.unmapped_probabilities
@@ -968,7 +970,7 @@ def test_output_regression(self):
                          MeanLearner()(self.housing), 1)
         out = self.get_output(widget.Outputs.predictions)
         np.testing.assert_equal(
-            out.metas,
+            out.metas[:, [0, 2]],
             np.hstack([pred.results.predicted.T for pred in widget.predictors]))
 
     def test_classless(self):
@@ -1188,6 +1190,43 @@ def test_migrate_shown_scores(self):
         self.widget.migrate_settings(settings, 1)
         self.assertTrue(settings["score_table"]["show_score_hints"]["Sensitivity"])
 
+    def test_output_error_reg(self):
+        data = self.housing
+        lin_reg = LinearRegressionLearner()
+        self.send_signal(self.widget.Inputs.data, data)
+        self.send_signal(self.widget.Inputs.predictors, lin_reg(data), 0)
+        self.send_signal(self.widget.Inputs.predictors,
+                         LinearRegressionLearner(fit_intercept=False)(data), 1)
+        pred = self.get_output(self.widget.Outputs.predictions)
+
+        names = ["", " (error)"]
+        names = [f"{n}{i}" for i in ("", " (1)") for n in names]
+        names = [f"{lin_reg.name}{x}" for x in names]
+        self.assertEqual(names, [m.name for m in pred.domain.metas])
+        self.assertAlmostEqual(pred.metas[0, 1], 6.0, 1)
+        self.assertAlmostEqual(pred.metas[0, 3], 5.1, 1)
+
+    def test_output_error_cls(self):
+        data = self.iris
+        log_reg = LogisticRegressionLearner()
+        self.send_signal(self.widget.Inputs.predictors, log_reg(data), 0)
+        self.send_signal(self.widget.Inputs.predictors,
+                         LogisticRegressionLearner(penalty="l1")(data), 1)
+        with data.unlocked(data.Y):
+            data.Y[1] = np.nan
+        self.send_signal(self.widget.Inputs.data, data)
+        pred = self.get_output(self.widget.Outputs.predictions)
+
+        names = [""] + [f" ({v})" for v in
+                        list(data.domain.class_var.values) + ["error"]]
+        names = [f"{n}{i}" for i in ("", " (1)") for n in names]
+        names = [f"{log_reg.name}{x}" for x in names]
+        self.assertEqual(names, [m.name for m in pred.domain.metas])
+        self.assertAlmostEqual(pred.metas[0, 4], 0.018, 3)
+        self.assertAlmostEqual(pred.metas[0, 9], 0.113, 3)
+        self.assertTrue(np.isnan(pred.metas[1, 4]))
+        self.assertTrue(np.isnan(pred.metas[1, 9]))
+
 
 class SelectionModelTest(unittest.TestCase):
     def setUp(self):