def command(job_id, split_desc, core_params, preprocessing_folder, model_folder, computation_parameters):
# LOADING INFO #
model_handler = PredictionModelInformationHandler(split_desc, core_params, preprocessing_folder, model_folder)
test_df = model_handler.get_test_df()
# COMPUTING SUBPOPULATION #
col_name = get_computation_parameter("column", computation_parameters)
col_type = get_type_of_column(col_name, model_handler)
if col_type == "CATEGORY":
value = get_computation_parameter("value", computation_parameters)
subpop_df = test_df[test_df[col_name] == value]
else:
raise NotImplementedError("Not implemented yet :-(")
# COMPUTING NEW METRICS ON SUBPOP #
prediction_type = model_handler.get_prediction_type()
if prediction_type == constants.BINARY_CLASSIFICATION:
results = compute_binary_subpopulation_metrics(subpop_df, model_handler)
else:
raise NotImplementedError("Not implemented yet :-(")
dkujson.dump_to_filepath(osp.join(model_folder, "subpop.json"), results)
return "ok"
def compute(job_id, split_desc, core_params, preprocessing_folder,
model_folder, computation_params):
if computation_params is None or "features_to_compute" not in computation_params:
raise Exception(
"'computation_params' should contains a key 'features_to_compute'")
model_handler = PredictionModelInformationHandler(split_desc, core_params,
preprocessing_folder,
model_folder)
features_to_compute = computation_params["features_to_compute"]
if model_handler.is_kfolding():
df = model_handler.get_full_df()
else:
df = model_handler.get_test_df()
progress = PartialDependenciesProgress(job_id, len(features_to_compute))
saver = PartialDependenciesSaver(model_folder, split_desc["schema"])
computer = PartialDependencyComputer(
df, model_handler.get_prediction_type(), model_handler.predict,
progress, model_handler.get_sample_weight_variable())
for index, feature_name in enumerate(features_to_compute):
drop_missing = model_handler.get_per_feature_col(feature_name).get(
"missing_handling") == "DROP_ROW"
feature_type = model_handler.get_type_of_column(feature_name)
is_dummified = False
category_possible_value = None
if feature_type == 'CATEGORY':
# nan values are replaced by a fake one because neither a scale nor a distribution can be computed with nan
feature_values = df[feature_name].fillna(
constants.FILL_NA_VALUE).values
is_dummified = model_handler.is_column_dummified(feature_name)
category_possible_value = model_handler.category_possible_values(
feature_name)
else:
feature_values = df[feature_name].values
pd_feature = PartialDependencyFeature(feature_type, feature_values,
feature_name, is_dummified,
category_possible_value,
drop_missing)
result = computer.compute(pd_feature)
saver.save(result)
progress.set_percentage((index + 1) * 100 / len(features_to_compute),
no_fail=False)