def test_should_contain_same_data(self):
# given
data = ClassifierData(are_samples_generated=False,
filename='datasets.xlsx',
number_of_dataset_if_not_generated=12)
# when
X1, y1 = ClassifLibrary.prepare_raw_data(data)
X2, y2 = ClassifLibraryOld.load_samples_from_datasets_first_two_rows(
classifier_data=ClassifierData(
number_of_dataset_if_not_generated=12))
# then
self.assertTrue(len(X2) == len(X1))
self.assertEqual(len(X1[0]), 2)
self.assertEqual(len(X2[0]), 2)
def test_should_return_right_number_of_subplots_when_external_plots_drawn(
self):
# given
data = ClassifierData(show_color_plot=True)
# when
target = ClassifLibrary.determine_number_of_subplots(data)
# then
self.assertEqual(self.NUMBER_OF_CLASSIFIERS * 2 + 1, target)
def test_should_return_one_permutation(self):
# given
generate_all_permutations = False
# when
classifier_data = ClassifierData(
generate_all_permutations=generate_all_permutations)
permutation = ClassifLibrary.generate_permutations(classifier_data)
# then
self.assertEqual(1, len(permutation))
self.assertEqual((0, 1), permutation[0])
def test_should_return_right_permutations(self):
# given
number_of_classifiers = 10
classifier_data = ClassifierData(
number_of_classifiers=number_of_classifiers)
# when
permutations = ClassifLibrary.generate_permutations(classifier_data)
# then
self.assertEqual(
len(list(permutations)),
int((number_of_classifiers + 2) * (number_of_classifiers + 1)))
def test_should_not_change_data(self):
# given
# when
X1, y1 = ClassifLibraryOld.load_samples_from_file_non_parametrized(
self.TEST_FILENAME)
X2, y2 = ClassifLibraryOld.load_samples_from_datasets_first_two_rows(
ClassifierData(number_of_dataset_if_not_generated=12))
# then
self.assertTrue(len(X2) <= len(X1))
for i in range(len(X2)):
self.assertTrue(X2[i] in X1)
def test_should_return_2_best_from_3_coefficients(self):
# given
coefficients = [[1, 2], [3, 4], [5, 6]]
scores = [[3], [1], [2]]
j = 0
classifier_data = ClassifierData()
# when
filtered_coefficients = ClassifLibrary.reduce_coefficients_in_subspace(
coefficients, scores, j, classifier_data)
# then
self.assertTrue([1, 2] in filtered_coefficients)
self.assertTrue([5, 6] in filtered_coefficients)
def test_should_select_right_features_when_swapped(self):
# given
X = [[0, 5, 10], [1, 0, 10], [2, 6, 10], [3, -1, 10], [4, 4, 10]]
y = [1, 0, 1, 0, 1]
expected_X0 = [[1, 0], [3, -1]]
expected_X1 = [[0, 5], [2, 6], [4, 4]]
classifier_data = ClassifierData(switch_columns_while_loading=True)
# when
X0, X1 = ClassifLibrary.make_selection(X, y, classifier_data)
# then
self.assertEqual(expected_X0, X0)
self.assertEqual(expected_X1, X1)
def test_should_contain_same_data_given_columns(self):
# given
data = ClassifierData(are_samples_generated=False,
filename='datasets.xlsx')
# when
X1, y1 = ClassifLibrary.prepare_raw_data(data)
X2, y2 = ClassifLibraryOld.load_samples_from_datasets_non_parametrised(
)
# then
self.assertTrue(len(X2) == len(X1))
self.assertEqual(len(X1[0]), 2)
self.assertEqual(len(X2[0]), 2)
def test_should_not_change_data_whole(self):
# given
data = ClassifierData(are_samples_generated=False,
filename='datasets.xlsx',
number_of_dataset_if_not_generated=12)
# when
X1, y1 = ClassifLibraryOld.load_samples_from_file_non_parametrized(
self.TEST_FILENAME)
X2, y2 = ClassifLibrary.prepare_raw_data(data)
# then
self.assertTrue(len(X2) <= len(X1))
self.assertEqual(len(X1[0]), 2)
self.assertEqual(len(X2[0]), 2)
def test_should_evaluate_weighted_average_coefficients_from_n_best(self):
# given
coefficients = [[1, 2], [3, 4], [5, 6], [7, 8]]
scores = [[0], [0.25], [0.5], [0.75]]
# when
a, b = \
ClassifLibrary.evaluate_weighted_average_coefficients_from_n_best(
coefficients, scores, 0,
ClassifierData(number_of_best_classifiers = 2, number_of_classifiers = len(scores)))
# then
self.assertEqual((coefficients[2][0] * scores[2][0] +
coefficients[3][0] * scores[3][0]) /
(scores[2][0] + scores[3][0]), a)
self.assertEqual((coefficients[2][1] * scores[2][0] +
coefficients[3][1] * scores[3][0]) /
(scores[2][0] + scores[3][0]), b)
def test_should_evaluate_average_coefficients_from_n_best(self):
# given
coefficients = [[1, 2], [3, 4], [5, 6], [7, 8], [9, 10]]
scores = [[0.25], [0], [0.5], [0.75], [1]]
# when
a, b = \
ClassifLibraryOld.evaluate_average_coefficients_from_n_best(
coefficients, scores, 0,
ClassifierData(number_of_best_classifiers = 3, number_of_classifiers = len(scores)))
# then
self.assertEqual(
(coefficients[2][0] + coefficients[3][0] + coefficients[4][0]) / 3,
a)
self.assertEqual(
(coefficients[2][1] + coefficients[3][1] + coefficients[4][1]) / 3,
b)
def test_should_return_3_best_from_5_coefficients(self):
# given
coefficients = [[1, 2], [3, 4], [5, 6], [7, 8], [9, 10]]
scores = [[3, 0], [1, 4], [2, 2], [4, 5], [5, 0]]
number_of_classifiers = 5
number_of_best_classifiers = 3
j = 1
classifier_data = ClassifierData(
number_of_classifiers=number_of_classifiers,
number_of_best_classifiers=number_of_best_classifiers)
# when
filtered_coefficients = ClassifLibrary.reduce_coefficients_in_subspace(
coefficients, scores, j, classifier_data)
# then
self.assertTrue([3, 4] in filtered_coefficients)
self.assertTrue([5, 6] in filtered_coefficients)
self.assertTrue([7, 8] in filtered_coefficients)
def test_should_return_right_dataset_permutation(self):
# given
X = [[0], [1], [2], [3], [4]]
y = [[5], [6], [7], [8], [9]]
tup = (1, 3)
# when
X_whole_train, y_whole_train, X_validation, y_validation, X_test, y_test = \
ClassifLibrary.get_permutation(X, y, tup, ClassifierData())
# then
self.assertEqual(X[tup[0]], X_validation)
self.assertEqual(y[tup[0]], y_validation)
self.assertEqual(X[tup[1]], X_test)
self.assertEqual(y[tup[1]], y_test)
for i in range(len(X)):
if i not in tup:
self.assertTrue(X[i] in X_whole_train)
self.assertTrue(y[i] in y_whole_train)