def test_fit_intercept_and_copy(coefs, intercept):
"""Test if fit_intercept and copy_X work."""
X, y = _create_dataset(coefs, intercept, noise=2.0)
imb = ImbalancedLinearRegression(fit_intercept=False, copy_X=False)
imb.fit(X, y)
assert imb.intercept_ == 0.0
def test_score(coefs, intercept):
"""Tests with noise on an easy problem. A good score should be possible."""
X, y = _create_dataset(coefs, intercept, noise=0.1)
imb = ImbalancedLinearRegression()
imb.fit(X, y)
assert imb.score(X, y) > 0.9
def test_over_estimation(coefs, intercept):
"""Test if the model is able to overestimate."""
X, y = _create_dataset(coefs, intercept, noise=2.0)
imb = ImbalancedLinearRegression(overestimation_punishment_factor=0.01)
imb.fit(X, y)
assert (imb.predict(X) < y).mean() < 0.15
def test_coefs_and_intercept__no_noise(coefs, intercept):
"""Regression problems without noise."""
X, y = _create_dataset(coefs, intercept)
imb = ImbalancedLinearRegression()
imb.fit(X, y)
assert imb.score(X, y) > 0.99
def test_coefs_and_intercept__no_noise_positive(coefs, intercept):
"""Test with only positive coefficients."""
X, y = _create_dataset(coefs, intercept)
imb = ImbalancedLinearRegression(positive=True)
imb.fit(X, y)
assert all(imb.coef_ >= 0)
assert imb.score(X, y) > 0.5
def test_coefs_and_intercept__no_noise_regularization(coefs, intercept):
"""Test model with regularization. The size of the coef vector should shrink the larger alpha gets."""
X, y = _create_dataset(coefs, intercept)
imbs = [
ImbalancedLinearRegression(alpha=alpha, l1_ratio=0.5).fit(X, y)
for alpha in range(4)
]
coef_size = np.array([np.sum(imb.coef_ ** 2) for imb in imbs])
for i in range(3):
assert coef_size[i] >= coef_size[i + 1]
def test_imbalanced(test_fn):
regr = ImbalancedLinearRegression()
test_fn(ImbalancedLinearRegression.__name__, regr)
@pytest.mark.parametrize("coefs, intercept", test_batch)
def test_fit_intercept_and_copy(coefs, intercept):
"""Test if fit_intercept and copy_X work."""
X, y = _create_dataset(coefs, intercept, noise=2.0)
imb = ImbalancedLinearRegression(fit_intercept=False, copy_X=False)
imb.fit(X, y)
assert imb.intercept_ == 0.0
@pytest.mark.parametrize("test_fn", [check_shape_remains_same_regressor])
def test_imbalanced(test_fn):
regr = ImbalancedLinearRegression()
test_fn(ImbalancedLinearRegression.__name__, regr)
@pytest.mark.parametrize(
"regr", [
(ImbalancedLinearRegression.__name__, ImbalancedLinearRegression()),
(ImbalancedLinearRegression.__name__ + "_positive", ImbalancedLinearRegression(positive=True)),
(ImbalancedLinearRegression.__name__ + "_positive__no_intercept", ImbalancedLinearRegression(positive=True, fit_intercept=False)),
(ImbalancedLinearRegression.__name__ + "_no_intercept", ImbalancedLinearRegression(fit_intercept=False))
]
)
@pytest.mark.parametrize(
"test_fn",
select_tests(
flatten([general_checks, nonmeta_checks, regressor_checks]),
)
)
def test_estimator_checks(regr, test_fn):
test_fn(*regr)