def test_regression_boston():
boston = load_boston()
data = data_df_from_bunch(boston)
er = SimpleRegressor()
er.fit(data, target_col='target')
# test nupmy array
er = SimpleRegressor()
er.fit(boston.data, boston.target)
def test_shuffle_cross_validation():
# somewhat nonlinear design with sorted target
rng = np.random.RandomState(42)
X = rng.normal(size=(100, 10))
w = rng.normal(size=(10, ))
y = np.dot(X, w)
y = .1 * y**2 + 2 * y
# throws off linear model if we sort
sorting = np.argsort(y)
X = pd.DataFrame(X[sorting, :])
y = pd.Series(y[sorting])
sr = SimpleRegressor(shuffle=False).fit(X, y)
assert sr.log_[-2].r2 < 0.1
sr = SimpleRegressor().fit(X, y)
assert sr.log_[-2].r2 > .9
def test_evaluate_score_ndim(X, y):
"""Test fit() works for both y.ndim == 1 and y.ndim == 2. Two test cases
are listed in @pytest.mark.parametrize()
"""
sr = SimpleRegressor(random_state=0)
print(f"Data ndim: X: {X.shape}, y: {y.shape}")
sr.fit(X, y)