deepinsight.doctor.posttraining package¶
Submodules¶
deepinsight.doctor.posttraining.model_information_handler module¶
-
class
deepinsight.doctor.posttraining.model_information_handler.PredictionModelInformationHandler(split_desc, core_params, preprocessing_folder, model_folder)¶ Bases:
object-
category_possible_values(col_name)¶ Get the list of modalities which are dummified by the preprocessing for the given column :param col_name: the name of the column :return: None if the column is not dummified else it returns the list of modalities that are dummified
-
get_clf()¶
-
get_collector_data()¶
-
get_full_df()¶
-
get_modeling_params()¶
-
get_per_feature()¶
-
get_per_feature_col(col_name)¶
-
get_pipeline(with_target=True)¶
-
get_posttrain_folder()¶
-
get_prediction_type()¶
-
get_predictor()¶
-
get_preproc_handler()¶
-
get_preprocessing_params()¶
-
get_role_of_column(col_name)¶
-
get_sample_weight_variable()¶
-
get_target_map()¶
-
get_target_variable()¶
-
get_test_df()¶
-
get_train_df()¶
-
get_type_of_column(col_name)¶
-
get_weight_method()¶
-
is_column_dummified(col_name)¶
-
is_ensemble()¶
-
is_keras_backend()¶
-
is_kfolding()¶
-
predict(df, output_probas=True)¶
-
predict_and_concatenate(df, output_probas=True)¶
-
prepare_for_scoring(df)¶
-
prepare_for_scoring_full(df)¶
-
run_binary_scoring(df, out_folder)¶
-
run_regression_scoring(df, out_folder)¶
-
set_keras_scoring_batches(new_value)¶
-
with_sample_weights()¶
-
deepinsight.doctor.posttraining.partial_depency module¶
-
class
deepinsight.doctor.posttraining.partial_depency.PartialDependenciesProgress(future_id, number_of_features)¶ Bases:
deepinsight.doctor.posttraining.percentage_progress.PercentageProgress-
set_percentage_for_single_computation(percentage, no_fail=True)¶
-
-
class
deepinsight.doctor.posttraining.partial_depency.PartialDependenciesSaver(folder, schema)¶ Bases:
object-
save(pd_result)¶
-
-
class
deepinsight.doctor.posttraining.partial_depency.PartialDependencyComputer(df, prediction_type, prediction_func, progress, sample_weights_col_name, max_cats=30)¶ Bases:
object-
aggregate_less_frequent_values(scale, distribution)¶
-
compute(pd_feature)¶
-
-
class
deepinsight.doctor.posttraining.partial_depency.PartialDependencyFeature(feature_type, values, name, is_dummified=False, dummified_modalities=None, drop_missing=False)¶ Bases:
object-
is_represented(value)¶ Returns True if the column is not dummified, else it checks if the value/modality is known by the model, e.g. if the preprocessing dummify this modality :param value: modality of the feature :return: boolean
-
-
class
deepinsight.doctor.posttraining.partial_depency.PartialDependencyResult(pd_feature, scale, distribution, partial_dependence, indices_to_drop=None, unrepresented_modalities=None)¶ Bases:
object
-
deepinsight.doctor.posttraining.partial_depency.compute(job_id, split_desc, core_params, preprocessing_folder, model_folder, computation_params)¶
deepinsight.doctor.posttraining.percentage_progress module¶
deepinsight.doctor.posttraining.subpopulation module¶
-
deepinsight.doctor.posttraining.subpopulation.command(job_id, split_desc, core_params, preprocessing_folder, model_folder, computation_parameters)¶
-
deepinsight.doctor.posttraining.subpopulation.compute_binary_subpopulation_metrics(subpop_df, model_handler)¶
-
deepinsight.doctor.posttraining.subpopulation.get_computation_parameter(key, computation_parameters)¶
-
deepinsight.doctor.posttraining.subpopulation.get_type_of_column(col_name, model_handler)¶