def testCN2Learner(self):
learner = CN2Learner()
# classic CN2 removes covered learning instances
self.assertTrue(learner.cover_and_remove ==
_RuleLearner.exclusive_cover_and_remove)
# Entropy measure is used to evaluate found hypotheses
self.assertTrue(
type(learner.rule_finder.quality_evaluator) == EntropyEvaluator)
# test that the learning requirements are relaxed by default
self.assertTrue(
learner.rule_finder.general_validator.max_rule_length >= 5)
self.assertTrue(
learner.rule_finder.general_validator.min_covered_examples == 1)
classifier = learner(self.titanic)
self.assertEqual(classifier.original_domain, self.titanic.domain)
# all learning instances are covered when limitations do not
# impose rule length or minimum number of covered examples
num_covered = np.sum(
[rule.curr_class_dist for rule in classifier.rule_list])
self.assertEqual(num_covered, self.titanic.X.shape[0])
# prediction (matrix-wise, all testing instances at once)
# test that returned result is of correct size
predictions = classifier.predict(self.titanic.X)
self.assertEqual(len(predictions), self.titanic.X.shape[0])
def setUp(self):
self.titanic = Table('titanic')
self.learner = CN2Learner()
self.classifier = self.learner(self.titanic)
# CN2Learner does not add `instances` attribute to the model, but
# the Rules widget does. We simulate the model we get from the widget.
self.classifier.instances = self.titanic
self.widget = self.create_widget(OWRuleViewer)
def setUpClass(cls):
super().setUpClass()
WidgetOutputsTestMixin.init(cls)
cls.titanic = Table('titanic')
cls.learner = CN2Learner()
cls.classifier = cls.learner(cls.titanic)
# CN2Learner does not add `instances` attribute to the model, but
# the Rules widget does. We simulate the model we get from the widget.
cls.classifier.instances = cls.titanic
cls.signal_name = "Classifier"
cls.signal_data = cls.classifier
cls.data = cls.titanic
def setUp(self):
self.titanic = Table('titanic')
self.learner = CN2Learner()
self.classifier = self.learner(self.titanic)
self.widget = self.create_widget(OWRuleViewer)
textoffset = pw + vmargin * 2
painter.save()
baseline = rect.bottom() - textoffset / 2
text = str(index.data(Qt.DisplayRole))
option.displayAlignment = Qt.AlignCenter
text_rect = rect.adjusted(0, 0, 0, -textoffset * 0)
self.drawDisplay(painter, option, text_rect, text)
painter.setRenderHint(QPainter.Antialiasing)
for prop, color in zip(curr_class_dist, self.color_schema):
if prop == 0:
continue
painter.setPen(QPen(QBrush(color), pw))
to_x = x + prop * width
line = QLineF(x, baseline, to_x, baseline)
painter.drawLine(line)
x = to_x
painter.restore()
painter.restore()
if __name__ == "__main__": # pragma: no cover
from Orange.classification import CN2Learner
data = Table("iris")
learner = CN2Learner()
model = learner(data)
model.instances = data
WidgetPreview(OWRuleViewer).run(model)