def test_threshold_encoder_with_all_columns_under_threshold():
encoder = ThresholdOneHotEncoder(threshold=0.01)
Xt = encoder.fit_transform(np.array([[1, 2], [1, 3]])).todense()
assert_array_equal(np.array([
[1, 1, 0],
[1, 0, 1],
]), Xt)
def test_threshold_encoder_with_no_columns_over_threshold():
encoder = ThresholdOneHotEncoder(threshold=1000)
Xt = encoder.fit_transform(np.array([[1, 2], [1, 3]])).todense()
assert_array_equal(np.array([
[0, 0],
[0, 0],
]), Xt)
def test_threshold_encoder(X, X_expected_categories, X_expected, max_categories, threshold):
enc = ThresholdOneHotEncoder(threshold=threshold, max_categories=max_categories)
X_observed_sparse = enc.fit_transform(X)
assert isinstance(X_observed_sparse, sp.csr_matrix)
assert len(enc.categories_) == len(X_expected_categories)
for observed_category, expected_category in zip(enc.categories_, X_expected_categories):
np.testing.assert_array_equal(observed_category, expected_category)
X_observed_dense = X_observed_sparse.toarray()
np.testing.assert_array_equal(X_observed_dense, X_expected)
