def test_bootstrap(self):
X = numpy.array([[0, 1], [0, 0]])
y = numpy.array(['H', 'T'])
dataset = Dataset(X, y)
bootstrap = dataset.bootstrap(1000)
self.assertEqual(bootstrap.X.shape[0], 1000)
self.assertEqual('H' in bootstrap.y, True) # this has a 10e-302ish chance of failing
def test_more_complicated_splitting(self):
# x1 < 0.25 => 'a'
# x1 >= 0.25, x2 = 0 => 'b'
# x1 < 0.50, x2 = 1 => 'c'
# x1 >= 0.50, x2 = 1 => 'd'
X = numpy.array([[0.2, 0], [0.01, 1], [0.15, 0], [0.232, 1], [0.173, 0], [0.263, 0], [0.671, 0], [0.9, 0], [0.387, 1], [0.482, 1], [0.632, 1], [0.892, 1]])
y = numpy.array(['a', 'a', 'a', 'a', 'a', 'b', 'b', 'b', 'c', 'c', 'a', 'a'])
dataset = Dataset(X, y, [NumericAttribute(0), CategoricalAttribute(1)])
splitter = dataset.best_single_attribute_splitter()
self.assertEqual(splitter.attribute.index, 0)
self.assertGreaterEqual(splitter.value, 0.23)
self.assertLess(splitter.value, 0.27)
subset1, subset2 = dataset.split_on(splitter).values()
numpy.testing.assert_array_equal(subset1.y, ['a', 'a', 'a', 'a', 'a'])
splitter2 = subset2.best_single_attribute_splitter()
self.assertEqual(splitter2.attribute.index, 1)
self.assertEqual(splitter2.value, 0)
subset21, subset22 = subset2.split_on(splitter2).values()
numpy.testing.assert_array_equal(subset22.y, ['b', 'b', 'b'])
splitter21 = subset21.best_single_attribute_splitter()
self.assertEqual(splitter21.attribute.index, 0)
self.assertGreaterEqual(splitter21.value, 0.482)
self.assertLess(splitter21.value, 0.632)
subset211, subset212 = subset21.split_on(splitter21).values()
numpy.testing.assert_array_equal(subset211.y, ['c', 'c'])
numpy.testing.assert_array_equal(subset212.y, ['a', 'a'])
def test_split_on(self):
X = numpy.array([[0, 1], [0, 0], [1, 0]])
y = numpy.array(['H', 'T', 'T'])
dataset = Dataset(X, y)
c0, c1 = dataset.attributes
split = dataset.split_on(IsEqualSplitter(c1, 0))
numpy.testing.assert_array_equal(split[0].X, numpy.array([[0, 1]]))
numpy.testing.assert_array_equal(split[1].X, numpy.array([[0, 0], [1, 0]]))
def test_entropy(self):
X = numpy.array([[0, 1], [0, 0]])
y = numpy.array(['H', 'T'])
dataset = Dataset(X, y)
c0, c1 = dataset.attributes
self.assertEqual(dataset.splitter_entropy(IsEqualSplitter(c0, 0)), 1)
self.assertEqual(dataset.splitter_entropy(IsEqualSplitter(c0, 1)), 1)
self.assertEqual(dataset.splitter_entropy(IsEqualSplitter(c1, 0)), 0)
self.assertEqual(dataset.splitter_entropy(IsEqualSplitter(c1, 1)), 0)
best_splitter = dataset.best_single_attribute_splitter()
self.assertEqual(best_splitter.attribute.index, 1)
self.assertEqual(best_splitter.value, 0)
def test_multitype_splitting(self):
# x1 < 0.5, x2 = 0 => 'Red'
# x1 < 0.5, x2 = 1 => 'Yellow'
# x1 >= .5 => 'Green'
X = numpy.array([[0.25, 0], [0.33, 0], [0.31, 1], [0.12, 1], [0.45, 0], [0.52, 0], [0.81, 0], [0.67, 1], [0.51, 1]])
y = numpy.array(['Red', 'Red', 'Yellow', 'Yellow', 'Red', 'Green', 'Green', 'Green', 'Green'])
dataset = Dataset(X, y, [NumericAttribute(0), CategoricalAttribute(1)])
splitter = dataset.best_single_attribute_splitter()
self.assertEqual(splitter.attribute.index, 0)
self.assertGreaterEqual(splitter.value, 0.45)
self.assertLess(splitter.value, 0.52)
subset1, subset2 = dataset.split_on(splitter).values()
subsplitter = subset1.best_single_attribute_splitter()
self.assertEqual(subsplitter.attribute.index, 1)
self.assertEqual(subsplitter.value, 0)
