def __init__(self, project: Project):
self.change_collector = ChangeCollector(project)
X_df, y_df = project.api.load_data()
encoder = project.api.encoder
y = encoder.fit_transform(y_df)
self.X_df, self.y = subsample_data_for_validation(X_df, y)
def validate_feature_acceptance(accepter_class, feature, features, X_df, y_df,
X_df_val, y_val, subsample):
if subsample:
X_df, y_df, X_df_val, y_val = subsample_data_for_validation(
X_df, y_df, X_df_val, y_val)
accepter = accepter_class(X_df, y_df, X_df_val, y_val, features, feature)
return accepter.judge()
def validate_feature_api(
feature: Feature,
X_df: pd.DataFrame,
y_df: Union[pd.DataFrame, pd.Series],
subsample: bool,
log_advice: bool = False,
) -> bool:
logger.debug(f'Validating feature {feature!r}')
if subsample:
X_df, y_df = subsample_data_for_validation(X_df, y_df)
valid, failures, advice = check_from_class(FeatureApiCheck, feature, X_df,
y_df)
if valid:
logger.info('Feature is valid')
else:
if log_advice:
logger.info(
'Feature is NOT valid; here is some advice for resolving the '
'feature API issues.')
for failure, advice_item in zip(failures, advice):
logger.info(f'{failure}: {advice_item}')
else:
logger.info(f'Feature is NOT valid; failures were {failures}')
return valid
def _load_validation_data(self, X_df: pd.DataFrame, y_df: pd.DataFrame,
subsample: bool):
if X_df is None or y_df is None:
_X_df, _y_df = self.api.load_data()
if X_df is None:
X_df = _X_df
if y_df is None:
y_df = _y_df
if subsample:
X_df, y_df = subsample_data_for_validation(X_df, y_df)
return X_df, y_df
def validate_feature_api(feature, X, y, subsample=False):
logger.debug('Validating feature {feature!r}'.format(feature=feature))
if subsample:
X, y = subsample_data_for_validation(X, y)
valid, failures = check_from_class(FeatureApiCheck, feature, X, y)
if valid:
logger.info('Feature is valid')
else:
logger.info('Feature is NOT valid; failures were {failures}'.format(
failures=failures))
return valid
def validate_feature_acceptance(feature, X, y, subsample=False, path=None,
package=None):
if package is not None:
project = Project(package)
elif path is not None:
project = Project.from_path(path)
else:
project = Project.from_cwd()
if subsample:
X, y = subsample_data_for_validation(X, y)
# build project
result = project.build(X, y)
# load accepter for this project
Accepter = _load_class(project, 'validation.feature_accepter')
accepter = Accepter(result.X_df, result.y, result.features, feature)
return accepter.judge()
def __init__(self, project):
self.change_collector = ChangeCollector(project)
X, y = project.load_data()
self.X, self.y = subsample_data_for_validation(X, y)
