def test_accuracy_metric_4_binary_classification(self):
# 100% correct
expected = np.array([0, 1, 1, 1, 0, 0, 1, 1, 1, 0]).reshape((-1, 1))
prediction = expected.copy()
score = acc_metric(expected, prediction)
self.assertEqual(1, score)
# 100% incorrect
prediction = (expected.copy() - 1) * -1
score = acc_metric(expected, prediction)
self.assertAlmostEqual(-1, score)
# Random
prediction = np.array([0, 0, 0, 0, 0, 1, 1, 1, 1, 1])
score = acc_metric(expected, prediction)
self.assertAlmostEqual(0, score)
def calculate_score(solution, prediction, task_type, metric, num_classes,
all_scoring_functions=False):
if task_type == MULTICLASS_CLASSIFICATION:
solution_binary = np.zeros((prediction.shape[0], num_classes))
for i in range(solution_binary.shape[0]):
label = solution[i]
solution_binary[i, label] = 1
solution = solution_binary
elif task_type in [BINARY_CLASSIFICATION, REGRESSION]:
if len(solution.shape) == 1:
solution = solution.reshape((-1, 1))
if task_type not in TASK_TYPES:
raise NotImplementedError(task_type)
# todo
# надо чинить libscores
scoring_func = getattr(classification_metrics, metric)
if solution.shape != prediction.shape:
raise ValueError('Solution shape %s != prediction shape %s' %
(solution.shape, prediction.shape))
if all_scoring_functions:
score = dict()
if task_type in REGRESSION_TASKS:
cprediction = sanitize_array(prediction)
score['a_metric'] = a_metric(solution, cprediction,
task=task_type)
score['r2_metric'] = r2_metric(solution, cprediction,
task=task_type)
else:
csolution, cprediction = normalize_array(solution,
prediction)
score['bac_metric'] = bac_metric(csolution, cprediction,
task=task_type)
score['auc_metric'] = auc_metric(csolution, cprediction,
task=task_type)
score['f1_metric'] = f1_metric(csolution, cprediction,
task=task_type)
score['pac_metric'] = pac_metric(csolution, cprediction,
task=task_type)
score['acc_metric'] = acc_metric(csolution, cprediction,
task=task_type)
else:
if task_type in REGRESSION_TASKS:
cprediction = sanitize_array(prediction)
score = scoring_func(solution, cprediction, task=task_type)
else:
csolution, cprediction = normalize_array(solution,
prediction)
score = scoring_func(csolution, cprediction, task=task_type)
return score
def test_accuracy_metric_4_multilabel_classification(self):
# 100% correct
expected = np.array([[0, 0, 1, 1, 0, 1, 0, 1, 0, 1],
[1, 1, 0, 0, 1, 0, 1, 0, 1, 0], [1, 1, 0, 0, 1, 0,
1, 0, 1, 0]])
prediction = expected.copy()
score = acc_metric(expected, prediction)
self.assertEqual(1, score)
# 100% incorrect
prediction = (expected.copy() - 1) * -1
score = acc_metric(expected, prediction)
self.assertAlmostEqual(-1, score)
# Pseudorandom
prediction = np.array([[0, 0, 0, 0, 0, 1, 1, 1, 1, 1], [0, 0, 0, 0, 0,
1, 1, 1, 1, 1],
[0, 0, 0, 0, 0, 1, 1, 1, 1, 1]])
score = acc_metric(expected, prediction)
self.assertAlmostEqual(-0.0666666666, score)
