def test_stores_subset(self):
path = datasetsDir(self) + 'numerical' + SEP + 'person'
_training, attributes, klass, _test, _gold = self.get_instances(path, True, False)
disc = discretise.Discretiser(_training, attributes, klass, _test, _gold, [4,6], [2,2])
self.assertEqual(2, len(disc.subset))
self.assertEqual(4, disc.subset[0].index)
self.assertEqual(6, disc.subset[1].index)
def test_option_cannot_be_zero(self):
path = datasetsDir(self) + 'numerical' + SEP + 'person'
try:
_training, attributes, klass, _test, _gold = self.get_instances(path, True, False)
disc = discretise.Discretiser(_training, attributes, klass, _test, _gold, [4,6], [2,0])
self.fail('should raise error as an option is zero')
except inv.InvalidDataError:
pass
def test_instances_attributes_and_options_are_extracted_from_strings(self):
path = datasetsDir(self) + 'numerical' + SEP + 'person'
_training, attributes, klass, _test, _gold = self.get_instances(path, True, False)
disc = discretise.Discretiser(_training, attributes, klass, _test, _gold, [0,1,4,5,6,7], [2,3,2,3,4,2])
self.assertEqual(6, len(disc.training))
self.assertEqual(2, len(disc.test))
self.assertEqual([0, 1, 4, 5, 6, 7], disc.attribute_indices)
self.assertEqual([2, 3, 2, 3, 4, 2], disc.options)
def test_returns_array_of_discretised_attributes(self):
path = datasetsDir(self) + 'numerical' + SEP + 'person'
_training, attributes, klass, _test, _gold = self.get_instances(path, True, False)
disc = discretise.Discretiser(_training, attributes, klass, _test, _gold, [4,6], [2,4])
disc_attrs = disc.discretised_attributes([nr.Range(0, 2), nr.Range(0, 120000)])
self.assertEqual(2, len(disc_attrs))
self.assertEqual(4, disc_attrs[0].index)
self.assertEqual(2, len(disc_attrs[0].values))
self.assertEqual(4, len(disc_attrs[1].values))
def test_naive_supervised_discretisation(self):
path = datasetsDir(self) + 'numerical' + SEP + 'person'
_training, attributes, klass, _test, _gold = self.get_instances(path, True, False)
disc = discretise.Discretiser(_training, attributes, klass, _test, _gold, [1])
self.assertEqual(1, len(disc.attributes[1].values))
disc.naive_supervised()
self.assertEqual(3, len(disc.attributes[1].values))
def test_unsupervised_equal_frequency(self):
path = datasetsDir(self) + 'numerical' + SEP + 'weather'
_training, attributes, klass, _test, _gold = self.get_instances(path)
disc = discretise.Discretiser(_training, attributes, klass, _test, _gold, [1], [3])
self.assertTrue(disc.attributes[1].is_continuous())
self.assertEqual(27.5, disc.training[0].value(disc.attributes[1]))
self.assertEqual(32, disc.training[2].value(disc.attributes[1]))
self.assertEqual(25.4, disc.test[0].value(disc.attributes[1]))
values = disc.training.values_grouped_by_attribute([disc.attributes[1]])
values[0].sort()
self.assertEqual([6.0, 9.0, 9.0, 10.699999999999999, 12.0, 12.0, 12.0, 14.1, 18.0, 27.5, 32.0, 33.100000000000001], values[0])
disc.unsupervised_equal_frequency()
self.assertFalse(disc.attributes[1].is_continuous())
self.assertEqual(4, len(disc.attributes[1].values))
self.assertEqual('c', disc.training[0].value(disc.attributes[1]))
self.assertEqual('d', disc.training[2].value(disc.attributes[1]))
self.assertEqual('c', disc.test[0].value(disc.attributes[1]))
def test_unsupervised_equal_width_discretisation(self):
path = datasetsDir(self) + 'numerical' + SEP + 'person'
_training, attributes, klass, _test, _gold = self.get_instances(path, True, False)
disc = discretise.Discretiser(_training, attributes, klass, _test, _gold, [1,4,5,6,7], [3,2,3,4,2])
self.assertTrue(disc.attributes[0].is_continuous())
self.assertTrue(disc.attributes[1].is_continuous())
self.assertTrue(disc.attributes[4].is_continuous())
self.assertTrue(disc.attributes[5].is_continuous())
self.assertTrue(disc.attributes[6].is_continuous())
self.assertTrue(disc.attributes[7].is_continuous())
self.assertEqual(25, disc.training[0].value(disc.attributes[1]))
self.assertEqual(26, disc.test[0].value(disc.attributes[1]))
disc.unsupervised_equal_width()
self.assertTrue(disc.attributes[0].is_continuous())
self.assertFalse(disc.attributes[1].is_continuous())
self.assertFalse(disc.attributes[4].is_continuous())
self.assertFalse(disc.attributes[5].is_continuous())
self.assertFalse(disc.attributes[6].is_continuous())
self.assertFalse(disc.attributes[7].is_continuous())
self.assertEqual('a', disc.training[0].value(disc.attributes[1]))
self.assertEqual('a', disc.test[0].value(disc.attributes[1]))
