def test_split_classification_many_imbalanced_classes(self):
for i in range(10):
X = np.array([range(20), range(20)]).transpose()
y = np.array((0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 2, 3,
4, 5))
np.random.shuffle(y)
X_train, X_valid, Y_train, Y_valid = split_data(
X, y,
classification=True)
print(X_train, Y_train)
self.assertLessEqual(max(Y_valid), 1)
def _split_regular(self):
X = np.array([[1, 2], [3, 4], [1, 2], [3, 4], [1, 2], [3, 4]])
y = np.array([0, 0, 0, 1, 1, 2])
X_train, X_valid, Y_train, Y_valid = split_data(X, y)
# Check shapes
self.assertEqual(X_train.shape, (4, 2))
self.assertEqual(Y_train.shape, (4, ))
self.assertEqual(X_valid.shape, (2, 2))
self.assertEqual(Y_valid.shape, (2, ))
self.assertListEqual(list(Y_valid), [0, 0])
self.assertListEqual(list(Y_train), [2, 0, 1, 1])
def _stratify(self):
X = np.array([[1, 2], [3, 4], [1, 2], [3, 4], [1, 2], [3, 4]])
y = np.array([0, 0, 0, 0, 1, 1])
X_train, X_valid, Y_train, Y_valid = split_data(X, y)
# Check shapes
self.assertEqual(X_train.shape[0], 4)
self.assertEqual(X_train.shape[1], 2)
self.assertEqual(Y_train.shape[0], 4)
self.assertEqual(X_valid.shape[0], 2)
self.assertEqual(X_valid.shape[1], 2)
self.assertEqual(Y_valid.shape[0], 2)
self.assertListEqual(list(Y_valid), [1, 0])
self.assertListEqual(list(Y_train), [0, 0, 0, 1])
def test_split_data_regression(self):
n_points = 1000
np.random.seed(42)
n_dims = np.random.randint(1, 100)
X = np.random.rand(n_points, n_dims)
y = np.random.rand(n_points)
X_train, X_valid, Y_train, Y_valid = split_data(X, y)
self.assertEqual(X_train.shape[0], 670)
self.assertEqual(X_valid.shape[0], 330)
self.assertEqual(Y_train.shape[0], 670)
self.assertEqual(Y_valid.shape[0], 330)
self.assertEqual(X_train.shape[1], n_dims)
self.assertEqual(X_valid.shape[1], n_dims)
# Random checks
self.assertAlmostEqual(X_train[4, 2], 0.5986584841970366)
self.assertAlmostEqual(X_valid[4, 2], 0.63911512838980322)
def _save_ensemble_data(x_data, y_data, tmp_dir, watcher):
"""Split dataset and store Data for the ensemble script.
:param x_data:
:param y_data:
:return:
"""
task_name = 'LoadData'
watcher.start_task(task_name)
_, _, _, y_ensemble = split_data(x_data, y_data)
filepath = os.path.join(tmp_dir, 'true_labels_ensemble.npy')
lock_path = filepath + '.lock'
with lockfile.LockFile(lock_path):
if not os.path.exists(filepath):
np.save(filepath, y_ensemble)
watcher.stop_task(task_name)
def __init__(self, data_manager, configuration,
with_predictions=False,
all_scoring_functions=False,
seed=1,
output_dir=None,
output_y_test=False,
num_run=None):
super(HoldoutEvaluator, self).__init__(
data_manager, configuration,
with_predictions=with_predictions,
all_scoring_functions=all_scoring_functions,
seed=seed,
output_dir=output_dir,
output_y_test=output_y_test,
num_run=num_run)
classification = data_manager.info['task'] in CLASSIFICATION_TASKS
self.X_train, self.X_optimization, self.Y_train, self.Y_optimization = \
split_data(data_manager.data['X_train'],
data_manager.data['Y_train'],
classification=classification)
self.model = self.model_class(self.configuration, self.seed)