def test_bin_transform(n_features=10, n_samples=10000):
"""
Testing BinTransformer
"""
data = numpy.random.normal(size=[n_samples, n_features])
n_bins = 41
transformer = BinTransformer(max_bins=n_bins).fit(data)
result = transformer.transform(data)
assert numpy.all(result < n_bins)
assert numpy.all(result >= 0)
assert numpy.allclose(numpy.max(result, axis=0), n_bins - 1)
assert numpy.allclose(numpy.min(result, axis=0), 0)
assure_monotonic(data, result)
# check reproducibility
assert numpy.all(
transformer.transform(data) == transformer.transform(data))
# checking dtype is integer
numpy_result = numpy.array(result)
print(numpy_result.dtype)
assert numpy_result.dtype == 'uint8'
def test_bin_transformer_extend_to(n_features=10, n_bins=123):
extended_length = 19
X, y = generate_sample(n_samples=20, n_features=n_features)
X1 = BinTransformer(max_bins=n_bins).fit(X).transform(X)
X2 = BinTransformer(max_bins=n_bins).fit(X).transform(
X, extend_to=extended_length)
assert len(X2) % extended_length == 0, 'wrong shape!'
assert numpy.allclose(X2[:len(X1)],
X1), 'extending does not work as expected!'
def test_network_with_scaler(n_samples=200, n_features=15, distance=0.5):
X, y = generate_sample(n_samples=n_samples, n_features=n_features, distance=distance)
for scaler in [BinTransformer(max_bins=16), IronTransformer()]:
clf = nnet.SimpleNeuralNetwork(scaler=scaler, epochs=300)
clf.fit(X, y)
p = clf.predict_proba(X)
assert roc_auc_score(y, p[:, 1]) > 0.8, 'quality is too low for model: {}'.format(clf)
def test_bin_transformer_limits(n_features=10, n_bins=123):
X, y = generate_sample(n_samples=1999, n_features=n_features)
X = BinTransformer(max_bins=n_bins).fit_transform(X)
assert numpy.allclose(X.max(axis=0), n_bins - 1)
X_orig, y = generate_sample(n_samples=20, n_features=n_features)
X = BinTransformer(max_bins=n_bins).fit_transform(X_orig)
assert numpy.allclose(X.min(axis=0), 0)
def test_bin_transform(n_features=10, n_samples=10000):
"""
Testing BinTransformer
"""
data = numpy.random.normal(size=[n_samples, n_features])
n_bins = 41
transformer = BinTransformer(max_bins=n_bins).fit(data)
result = transformer.transform(data)
assert numpy.all(result < n_bins)
assert numpy.all(result >= 0)
assert numpy.allclose(numpy.max(result, axis=0), n_bins - 1)
assert numpy.allclose(numpy.min(result, axis=0), 0)
assure_monotonic(data, result)
# check reproducibility
assert numpy.all(transformer.transform(data) == transformer.transform(data))
# checking dtype is integer
numpy_result = numpy.array(result)
print(numpy_result.dtype)
assert numpy_result.dtype == 'uint8'
def test_preprocessing(self):
X = np.array([[1.1, 1.2, 1.3],[5.1, 6.4, 10.5]])
transformer = BinTransformer().fit(X)
new_X = transformer.transform(X)
self.assertEqual((2, 3), new_X.shape)