def setUp(self):
self.meta = simple_meta_attrs()
self.samples = [Sample([i, 0], self.meta, last_is_class=True)
for i in xrange(10)] + \
[Sample([i, 1], self.meta, last_is_class=True)
for i in xrange(10, 15)]
self.classifiers = tuple(StupidClassifer(i) for i in xrange(2))
self.basics = tuple([(10., 0., 5., 0.), (0., 5., 0., 10.)])
def setUp(self):
self.N = 100
self.samples = [
Sample([1, 2, 3, 4]),
Sample([1, 1, 1, 1]),
Sample([2, 2, 2, 2]),
Sample([3, 3, 3, 3])
]
self.ru = [ReactiveUnit(s.get_values()) for s in self.samples]
def test_classify_similar(self):
self._init()
samples = [
Sample([0.9, 0.9, 0.9, 0.9]),
Sample([1.1, 1.1, 1.1, 1.1]),
Sample([0.9, 1.1, 0.9, 1.1]),
Sample([1, 1, 1, 1])
]
wc = [self.sc.classify(s) for s in samples]
wc.sort()
self.assertEqual(wc[0], wc[len(wc) - 1])
def test_classification(self):
cls_meta = NominalAttribute([0, 1])
meta = [NumericAttribute() for _ in xrange(3)]
train_set = [
Sample([0, 0, 0], meta, 0, cls_meta),
Sample([0, 1, 0], meta, 0, cls_meta),
Sample([0, 0, 1], meta, 0, cls_meta),
Sample([3, 0, 0], meta, 1, cls_meta),
Sample([3, 1, 0], meta, 1, cls_meta),
Sample([3, 0, 1], meta, 1, cls_meta),
]
classifier = self.knn1
# classifier = self.tree
classifier.train(train_set)
expected = [str(x) for x in [0, 0, 0, 1, 1, 1, 0, 1]]
samples = [s for s in train_set]
samples.extend([Sample([1, 0, 0], meta), Sample([2, 1, 0], meta)])
for e, s in izip(expected, samples):
self.assertEqual(e, classifier.classify(s))
k, p = classifier.classify_pval(s)
self.assertTrue(0. <= p <= 1.)
self.assertEqual(k, classifier.classify(s))
p2 = classifier.class_probabilities(s)
self.assertAlmostEqual(1., sum(p2.values()), delta=0.00001)
def test_classification(self):
classifier = self.knn1
# classifier = self.tree
classifier.train(self.train_set)
expected = map(str, [0, 0, 0, 1, 1, 1, 0, 1])
samples = [s for s in self.train_set]
samples.extend(
[Sample([1, 0, 0], self.meta),
Sample([2, 1, 0], self.meta)])
for e, s in izip(expected, samples):
self.assertEqual(e, classifier.classify(s))
k, p = classifier.classify_pval(s)
self.assertTrue(0. <= p <= 1.)
self.assertEqual(k, classifier.classify(s))
p2 = classifier.class_probabilities(s)
self.assertAlmostEqual(1., sum(p2.values()), delta=0.00001)
def test_equality(self):
self.assertNotEqual(self.sample, self.sample_cl)
meta = [NumericAttribute(), NominalAttribute(animals)]
sample = Sample([1.2, meta[1].get_idx("dog")], meta)
self.assertEqual(self.sample, sample)
self.assertNotEqual(self.sample, self.sample_cl)
meta = [NumericAttribute(), NominalAttribute(animals), NumericAttribute()]
sample = Sample([1.2, meta[1].get_idx("dog"), 3.14], meta)
self.assertNotEqual(self.sample, sample)
self.assertNotEqual(self.sample_cl, sample)
meta = [NumericAttribute(), NominalAttribute(animals)]
sample = Sample([1.2, meta[1].get_idx("cat")], meta)
self.assertNotEqual(self.sample, sample)
self.assertNotEqual(self.sample_cl, sample)
sample = Sample([1.3, meta[1].get_idx("dog")], meta)
self.assertNotEqual(self.sample, sample)
self.assertNotEqual(self.sample_cl, sample)
sample = Sample([100, self.meta[1].get_idx("cat")], self.meta,
self.meta_cl.get_idx("duck"), self.meta_cl)
self.assertEqual(self.sample_cl, sample)
self.assertNotEqual(self.sample, sample)
sample = Sample([10.20, self.meta[1].get_idx("cat")], self.meta,
self.meta_cl.get_idx("duck"), self.meta_cl)
self.assertNotEqual(self.sample, sample)
self.assertNotEqual(self.sample_cl, sample)
def setUp(self):
self.meta = simple_meta_attrs(['-', '+'])
self.cs = lambda i, v: Sample(
[i, self.meta[1].set_value(v)], self.meta, last_is_class=True)
self.classifier = OrangeClassifier('kNNLearner', k=1)
test_samples = '+++-++-+-+--+---'
N = len(test_samples)
train_samples = ('+' * (N / 2)) + ('-' * (N / 2))
self.test_samples, self.train_samples = ([
self.cs(i, v) for i, v in enumerate(samples)
] for samples in [test_samples, train_samples])
random.shuffle(self.test_samples)
self.classifier.train(self.train_samples)
def test_random_stimuli_with_distance(self):
samples = [Sample([x]) for x in xrange(10)] * 10
chooser = RandomStimuliChooser(None, True, 3)
env = Environment(samples, chooser)
for _ in xrange(10):
sort = sorted([x.get_values()[0] for x in env.get_stimuli(4)])
self.assertEqual([0, 3, 6, 9], sort)
chooser = RandomStimuliChooser(None, True, 5)
env = Environment(samples, chooser)
for _ in xrange(10):
sort = sorted([x.get_values()[0] for x in env.get_stimuli(2)])
self.assertEqual(len(sort), 2)
self.assertTrue(sort[0] < 5)
self.assertTrue(sort[1] >= 5)
self.assertRaises(Exception, chooser.get_stimuli, samples, 100)
def setUp(self):
self.classifiers = [("BayesLearner", [], {}), ("TreeLearner", [], {}),
("kNNLearner", [], {
"k": 1
}), ("kNNLearner", [], {
"k": 3
}), ("TreeLearner", [], {})]
self.knn1 = OrangeClassifier(self.classifiers[2][0])
self.knn3 = OrangeClassifier(self.classifiers[3][0])
self.tree = OrangeClassifier(self.classifiers[4][0])
self.cls_meta = NominalAttribute([0, 1])
self.meta = [NumericAttribute() for _ in xrange(3)]
self.train_set = [
Sample([0, 0, 0], self.meta, 0, self.cls_meta),
Sample([0, 1, 0], self.meta, 0, self.cls_meta),
Sample([0, 0, 1], self.meta, 0, self.cls_meta),
Sample([3, 0, 0], self.meta, 1, self.cls_meta),
Sample([3, 1, 0], self.meta, 1, self.cls_meta),
Sample([3, 0, 1], self.meta, 1, self.cls_meta),
]
def setUp(self):
self.meta = [NumericAttribute(), NominalAttribute(animals)]
self.sample = Sample([1.2, self.meta[1].get_idx("dog")], self.meta)
self.meta_cl = NominalAttribute(animals)
self.sample_cl = Sample([100, self.meta[1].get_idx("cat")], self.meta,
self.meta_cl.get_idx("duck"), self.meta_cl)