def test_ValueError_decode(self):
try:
encoder = Encoder(df=self.df, encoding=self.encoding)
encoder.decode(df=self.df,
encoding=create_test_encoding(data_encoding='None'))
except ValueError:
pass
def test_encode(self):
encoder = Encoder(df=self.df, encoding=self.encoding)
encoded_df = self.df.copy()
encoder.encode(df=encoded_df, encoding=self.encoding)
self.assertDictEqual(self.how_it_should_be.to_dict(),
encoded_df.to_dict())
def test_repeated_encode_decode(self):
encoder = Encoder(df=self.df, encoding=self.encoding)
encoded_df = self.df.copy()
encoder.encode(df=encoded_df, encoding=self.encoding)
encoder.decode(df=encoded_df, encoding=self.encoding)
encoder.encode(df=encoded_df, encoding=self.encoding)
encoder.decode(df=encoded_df, encoding=self.encoding)
self.assertDictEqual(self.df.to_dict(), encoded_df.to_dict())
def test_NotImplementedException_decode(self):
try:
encoder = Encoder(df=self.df, encoding=self.encoding)
encoder.decode(
df=self.df,
encoding=create_test_encoding(
data_encoding=DataEncodings.ONE_HOT_ENCODER.value))
except NotImplementedError:
pass
def retrieve_proper_encoder(job: Job) -> Encoder:
if job.incremental_train is not None:
return retrieve_proper_encoder(job.incremental_train)
else:
training_log, test_log, additional_columns = get_train_test_log(job.split)
training_df, _ = encode_label_logs(training_log, test_log, job, additional_columns=additional_columns,
encode=False)
return Encoder(training_df, job.encoding)
def _data_encoder_encoder(job: Job, training_log, test_log) -> None:
if job.type != JobTypes.LABELLING.value and \
job.encoding.value_encoding != ValueEncodings.BOOLEAN.value and \
job.predictive_model.predictive_model != PredictiveModels.TIME_SERIES_PREDICTION.value:
if job.incremental_train is not None:
encoder = retrieve_proper_encoder(job.incremental_train)
else:
if job.predictive_model.predictive_model != PredictiveModels.TIME_SERIES_PREDICTION.value and \
job.predictive_model.predictive_model != PredictiveModels.REGRESSION.value:
encoder = Encoder(training_log, job.encoding)
elif job.predictive_model.predictive_model == PredictiveModels.REGRESSION.value:
encoder = Encoder(training_log.drop('label', axis=1), job.encoding)
encoder.encode(training_log, job.encoding)
encoder.encode(test_log, job.encoding)
def encode_label_logs(training_log: EventLog,
test_log: EventLog,
job: Job,
additional_columns=None):
training_log, cols = _encode_log(training_log,
job.encoding,
job.labelling,
additional_columns=additional_columns,
cols=None)
# TODO pass the columns of the training log
print('\tDataset not found in cache, building..')
test_log, _ = _encode_log(test_log,
job.encoding,
job.labelling,
additional_columns=additional_columns,
cols=cols)
labelling = job.labelling
if (labelling.threshold_type in [
ThresholdTypes.THRESHOLD_MEAN.value,
ThresholdTypes.THRESHOLD_CUSTOM.value
]) and (labelling.type in [
LabelTypes.ATTRIBUTE_NUMBER.value, LabelTypes.DURATION.value,
LabelTypes.REMAINING_TIME.value
]):
if labelling.threshold_type == ThresholdTypes.THRESHOLD_MEAN.value:
threshold = training_log['label'].astype(float).mean()
elif labelling.threshold_type == ThresholdTypes.THRESHOLD_CUSTOM.value:
threshold = float(labelling.threshold)
else:
threshold = -1
training_log['label'] = training_log['label'].astype(float) < threshold
test_log['label'] = test_log['label'].astype(float) < threshold
if job.type != JobTypes.LABELLING.value and job.encoding.value_encoding != ValueEncodings.BOOLEAN.value and \
job.predictive_model.predictive_model != PredictiveModels.TIME_SERIES_PREDICTION.value:
# init nominal encode
encoder = Encoder(training_log, job.encoding)
encoder.encode(training_log, job.encoding)
encoder.encode(test_log, job.encoding)
return training_log, test_log
def test_ValueError_init_encoder(self):
try:
Encoder(df=self.df,
encoding=create_test_encoding(data_encoding='None'))
except ValueError:
pass
def test_encoder(self):
encoder = Encoder(df=self.df, encoding=self.encoding)
self.assertIsNotNone(encoder._encoder)
self.assertIsNotNone(encoder._label_dict)
self.assertIsNotNone(encoder._label_dict_decoder)