def fit(self, data):
LOGGER.info("Start scale data fit ...")
scale_value_results = []
self.header = data.schema.get('header')
if self.scale_param.method == consts.MINMAXSCALE:
min_max_scaler = MinMaxScaler(mode=self.scale_param.mode, area=self.scale_param.area,
feat_upper=self.scale_param.feat_upper,
feat_lower=self.scale_param.feat_lower,
out_upper=self.scale_param.out_upper, out_lower=self.scale_param.out_lower)
data, cols_scale_value = min_max_scaler.fit(data)
scale_value_results.append(cols_scale_value)
self.cols_scale_value = cols_scale_value
elif self.scale_param.method == consts.STANDARDSCALE:
standard_scaler = StandardScaler(with_mean=self.scale_param.with_mean, with_std=self.scale_param.with_std)
data, mean, std = standard_scaler.fit(data)
scale_value_results.append(mean)
scale_value_results.append(std)
self.mean = mean
self.std = std
else:
LOGGER.info("Scale method is {}, do nothing and return!".format(self.scale_param.method))
data.schema['header'] = self.header
LOGGER.info("End fit data ...")
return data, scale_value_results
def test_transform1(self):
standard_scaler = StandardScaler(with_mean=True, with_std=True)
fit_instance, mean, std = standard_scaler.fit(self.table_instance)
transform_data = standard_scaler.transform(self.table_instance, mean,
std)
self.assertListEqual(self.get_table_instance_feature(transform_data),
self.get_table_instance_feature(fit_instance))
def test_cols_select_fit_and_transform_repeat(self):
scale_column_idx = [1, 1, 2, 2, 4, 5, 5]
standard_scaler = StandardScaler(area='col',
scale_column_idx=scale_column_idx,
with_mean=True,
with_std=True)
fit_data, scale_conf = standard_scaler.fit(self.table_instance)
mean, std, scale_column_idx = scale_conf[0], scale_conf[1], scale_conf[
2]
scaler = SSL(with_mean=True, with_std=True)
scaler.fit(self.test_data)
transform_data = np.around(scaler.transform(self.test_data),
4).tolist()
for i, line in enumerate(transform_data):
for j, cols in enumerate(line):
if j not in scale_column_idx:
transform_data[i][j] = self.test_data[i][j]
self.assertListEqual(self.get_table_instance_feature(fit_data),
transform_data)
std_scale_transform_data = standard_scaler.transform(
self.table_instance, mean, std, scale_column_idx)
self.assertListEqual(
self.get_table_instance_feature(std_scale_transform_data),
transform_data)
def test_transform4(self):
standard_scaler = StandardScaler(with_mean=False, with_std=False)
fit_instance, scale_conf = standard_scaler.fit(self.table_instance)
mean, std, scale_column_idx = scale_conf[0], scale_conf[1], scale_conf[
2]
transform_data = standard_scaler.transform(self.table_instance, mean,
std, scale_column_idx)
self.assertListEqual(self.get_table_instance_feature(transform_data),
self.get_table_instance_feature(fit_instance))
def test_fit4(self):
standard_scaler = StandardScaler(with_mean=False, with_std=False)
fit_instance, mean, std = standard_scaler.fit(self.table_instance)
scaler = SSL(with_mean=False, with_std=False)
scaler.fit(self.test_data)
self.assertListEqual(
self.get_table_instance_feature(fit_instance),
np.around(scaler.transform(self.test_data), 4).tolist())
self.assertEqual(mean, [0 for _ in range(len(self.test_data[0]))])
self.assertEqual(std, [1 for _ in range(len(self.test_data[0]))])
def test_fit3(self):
standard_scaler = StandardScaler(with_mean=True, with_std=False)
fit_instance, mean, std = standard_scaler.fit(self.table_instance)
scaler = SSL(with_std=False)
scaler.fit(self.test_data)
self.assertListEqual(
self.get_table_instance_feature(fit_instance),
np.around(scaler.transform(self.test_data), 4).tolist())
self.assertListEqual(list(np.around(mean, 4)),
list(np.around(scaler.mean_, 4)))
self.assertListEqual(list(np.around(std, 4)), [1 for _ in std])
def test_fit2(self):
standard_scaler = StandardScaler(with_mean=False, with_std=True)
fit_instance, scale_conf = standard_scaler.fit(self.table_instance)
mean, std = scale_conf[0], scale_conf[1]
scaler = SSL(with_mean=False)
scaler.fit(self.test_data)
self.assertListEqual(
self.get_table_instance_feature(fit_instance),
np.around(scaler.transform(self.test_data), 4).tolist())
self.assertListEqual(list(np.around(mean, 4)), [0 for _ in mean])
self.assertListEqual(list(np.around(std, 4)),
list(np.around(scaler.scale_, 4)))
def test_fit6(self):
standard_scaler = StandardScaler(area='col',
with_mean=True,
with_std=True)
fit_instance, scale_conf = standard_scaler.fit(self.table_instance)
mean, std, scale_column_idx = scale_conf[0], scale_conf[1], scale_conf[
2]
scaler = SSL()
scaler.fit(self.test_data)
self.assertListEqual(
self.get_table_instance_feature(fit_instance),
np.around(scaler.transform(self.test_data), 4).tolist())
self.assertListEqual(list(np.around(mean, 4)),
list(np.around(scaler.mean_, 4)))
self.assertListEqual(list(np.around(std, 4)),
list(np.around(scaler.scale_, 4)))
def fit(self, data):
"""
Apply scale for input data
Parameters
----------
data: data_instance, input data
Returns
----------
data:data_instance, data after scale
scale_value_results: list, the fit results information of scale
"""
LOGGER.info("Start scale data fit ...")
self.header = data.schema.get('header')
if self.scale_param.method == consts.MINMAXSCALE:
min_max_scaler = MinMaxScaler(mode=self.scale_param.mode, area=self.scale_param.area,
scale_column_idx=self.scale_param.scale_column_idx,
feat_upper=self.scale_param.feat_upper,
feat_lower=self.scale_param.feat_lower,
out_upper=self.scale_param.out_upper, out_lower=self.scale_param.out_lower)
data, cols_scale_value = min_max_scaler.fit(data)
self.cols_scale_res = cols_scale_value
elif self.scale_param.method == consts.STANDARDSCALE:
standard_scaler = StandardScaler(area=self.scale_param.area,
scale_column_idx=self.scale_param.scale_column_idx,
with_mean=self.scale_param.with_mean, with_std=self.scale_param.with_std)
data, cols_scale_value = standard_scaler.fit(data)
self.mean = cols_scale_value[0]
self.std = cols_scale_value[1]
self.std_scale_column_idx = cols_scale_value[2]
self.cols_scale_res = cols_scale_value
else:
LOGGER.info("Scale method is {}, do nothing and return!".format(self.scale_param.method))
data.schema['header'] = self.header
LOGGER.info("End fit data ...")
return data, self.cols_scale_res