def test_polynomial_model():
#Lars excluded as it performs poorly.
polynomial_report = polyr(diabetes_data,
diabetes_target,
n_folds=2,
num_degrees=3,
verbose=1,
concurrency=1,
feature_selection=False,
save=False,
project_name='polynomial_test')
assert ((polynomial_report.scores.mean()[:, 'test']).median() > 0.3), \
'test score below chance'
def test_run_regression():
global report
report = polyr(diabetes_data,
diabetes_target,
n_folds=2,
verbose=1,
concurrency=1,
feature_selection=False,
scoring='r2',
save=False,
project_name='test_regression')
assert ((report.scores.mean()[:, 'test']).median() > 0.2),\
'test score below chance'
assert ((report.scores.mean()[:, 'train']).median() > 0.2),\
'train score below chance'
def test_feature_selection_regression():
global report_with_features
report_with_features = polyr(diabetes_data,
diabetes_target,
n_folds=2,
verbose=1,
concurrency=1,
feature_selection=True,
scoring='r2',
save=False,
project_name='test_feature_selection')
assert ((report_with_features.scores.mean()[:, 'test']).median() > 0.2),\
'test score below chance'
assert ((report_with_features.scores.mean()[:, 'train']).median() > 0.2),\
'train score below chance'
for key, ypred in report_with_features.predictions.iteritems():
mse = np.linalg.norm(ypred - diabetes_target) / len(diabetes_target)
assert mse < 5, '{} Prediction error is too high'.format(key)