def test_not_normalized(self, dummy_data, monkeypatch):
X, y, coef = dummy_data
selector = ForwardSelector(normalize=False)
mock_scaler = Mock()
mock_scaler.fit_transform.return_value = X
monkeypatch.setattr(selector, 'scaler', mock_scaler)
final = selector.fit_transform(X, y)
mock_scaler.fit_transform.assert_not_called()
def test_features_in(self, metric, n_params, dummy_train_test):
X_train, X_test, y_train, y_test = dummy_train_test
pl = Pipeline([('scaler', StandardScaler()),
('selector', ForwardSelector(metric=metric)),
('lr', LinearRegression())])
pl.fit(X_train, y_train)
expected_params = n_params
actual_lr_features = pl['lr'].n_features_in_
assert expected_params == actual_lr_features
def test_selected_params(self, metric, n_params, dummy_train_test):
X_train, X_test, y_train, y_test = dummy_train_test
pl = Pipeline([('scaler', StandardScaler()),
('selector', ForwardSelector(metric=metric)),
('lr', LinearRegression())])
pl.fit(X_train, y_train)
expected_params = n_params
actual_params = pl['selector'].get_support().sum()
assert actual_params == expected_params
def test_not_fitted(self):
selector = ForwardSelector()
with pytest.raises(NotFittedError):
selector.get_support()
def test_fit_one_params(self, dummy_data_one_relevant):
X, y, coef = dummy_data_one_relevant
selector = ForwardSelector()
final = selector.fit_transform(X, y)
np.testing.assert_equal(coef * selector.best_support_, coef)
def test_fit_bic(self, dummy_data):
X, y, coef = dummy_data
selector = ForwardSelector(metric='bic')
final = selector.fit_transform(X, y)
np.testing.assert_equal(coef * selector.best_support_, coef)