def test_transform_ts_files():
n_class1 = 4
n_class2 = 8
transform_type = "Train/Test Split"
time_series = [TimeSeries(*sample_values(), target='class1')
for i in range(n_class1)]
time_series += [TimeSeries(*sample_values(), target='class2')
for i in range(n_class2)]
output = transformation.transform_ts_files(time_series, transform_type)
npt.assert_equal(len(output), 2)
def test_train_test_split():
# Mock out unevenly-labeled test data: 4 class1, 8 class2
n_class1 = 4
n_class2 = 8
TS_MOCK_PATHS = [TS_PATHS[0]] * n_class1 + [TS_PATHS[1]] * n_class2
transform_type = "Train/Test Split"
time_series = [tslib.from_netcdf(path) for path in TS_MOCK_PATHS]
np.random.seed(0)
train, test = transformation.transform_ts_files(time_series,
transform_type)
npt.assert_equal(sum(ts.target == 'class1' for ts in train), 1 * n_class1 / 2)
npt.assert_equal(sum(ts.target == 'class1' for ts in test), n_class1 / 2)
npt.assert_equal(sum(ts.target == 'class2' for ts in train), 1 * n_class2 / 2)
npt.assert_equal(sum(ts.target == 'class2' for ts in test), n_class2 / 2)
def test_train_test_split():
# Mock out unevenly-labeled test data: 4 class1, 8 class2
n_class1 = 4
n_class2 = 8
transform_type = "Train/Test Split"
time_series = [TimeSeries(*sample_values(), target='class1')
for i in range(n_class1)]
time_series += [TimeSeries(*sample_values(), target='class2')
for i in range(n_class2)]
np.random.seed(0)
train, test = transformation.transform_ts_files(time_series,
transform_type)
npt.assert_equal(
sum(ts.target == 'class1' for ts in train), 1 * n_class1 / 2)
npt.assert_equal(sum(ts.target == 'class1' for ts in test), n_class1 / 2)
npt.assert_equal(
sum(ts.target == 'class2' for ts in train), 1 * n_class2 / 2)
npt.assert_equal(sum(ts.target == 'class2' for ts in test), n_class2 / 2)
