Merge pull request #1720 from janezd/fix-confusion-indices

[FIX] Confusion matrix: Map annotated data through row_indices, add probabi…

Author: ajdapretnar

Orange/widgets/evaluate/owconfusionmatrix.py

@@ -229,7 +229,7 @@ def set_results(self, results):
 
         data = None
         if results is not None and results.data is not None:
-            data = results.data
+            data = results.data[results.row_indices]
 
         if data is not None and not data.domain.has_discrete_class:
             self.Error.no_regression()
@@ -375,9 +375,8 @@ def commit(self):
             data.name = learner_name
 
             if selected:
-                row_indices = self.results.row_indices[selected]
-                annotated_data = create_annotated_table(data, row_indices)
-                data = data[row_indices]
+                annotated_data = create_annotated_table(data, selected)
+                data = data[selected]
             else:
                 annotated_data = create_annotated_table(data, [])
                 data = None

Orange/widgets/evaluate/tests/test_owconfusionmatrix.py

@@ -1,9 +1,11 @@
 # pylint: disable=missing-docstring
+import numpy as np
 
 from Orange.data import Table
 from Orange.classification import NaiveBayesLearner, TreeLearner
 from Orange.regression import MeanLearner
-from Orange.evaluation.testing import CrossValidation, TestOnTrainingData
+from Orange.evaluation.testing import CrossValidation, TestOnTrainingData, \
+    ShuffleSplit
 from Orange.widgets.evaluate.owconfusionmatrix import OWConfusionMatrix
 from Orange.widgets.tests.base import WidgetTest, WidgetOutputsTestMixin
 
@@ -16,11 +18,11 @@ def setUpClass(cls):
 
         bayes = NaiveBayesLearner()
         tree = TreeLearner()
-        iris = cls.data
+        cls.iris = cls.data
         titanic = Table("titanic")
         common = dict(k=3, store_data=True)
-        cls.results_1_iris = CrossValidation(iris, [bayes], **common)
-        cls.results_2_iris = CrossValidation(iris, [bayes, tree], **common)
+        cls.results_1_iris = CrossValidation(cls.iris, [bayes], **common)
+        cls.results_2_iris = CrossValidation(cls.iris, [bayes, tree], **common)
         cls.results_2_titanic = CrossValidation(titanic, [bayes, tree],
                                                 **common)
 
@@ -59,8 +61,7 @@ def _select_data(self):
     def test_show_error_on_regression(self):
         """On regression data, the widget must show error"""
         housing = Table("housing")
-        results = TestOnTrainingData(housing, [MeanLearner()])
-        results.data = housing
+        results = TestOnTrainingData(housing, [MeanLearner()], store_data=True)
         self.send_signal("Evaluation Results", results)
         self.assertTrue(self.widget.Error.no_regression.is_shown())
         self.send_signal("Evaluation Results", None)
@@ -69,3 +70,15 @@ def test_show_error_on_regression(self):
         self.assertTrue(self.widget.Error.no_regression.is_shown())
         self.send_signal("Evaluation Results", self.results_1_iris)
         self.assertFalse(self.widget.Error.no_regression.is_shown())
+
+    def test_row_indices(self):
+        """Map data instances when using random shuffling"""
+        results = ShuffleSplit(self.iris, [NaiveBayesLearner()],
+                               store_data=True)
+        self.send_signal("Evaluation Results", results)
+        self.widget.select_correct()
+        selected = self.get_output("Selected Data")
+        correct = np.equal(results.actual, results.predicted)[0]
+        correct_indices = results.row_indices[correct]
+        self.assertSetEqual(set(self.iris[correct_indices].ids),
+                            set(selected.ids))