streamline.postanalysis.model_replicate module

class streamline.postanalysis.model_replicate.ReplicateJob(dataset_filename, dataset_for_rep, full_path, class_label, instance_label, match_label, ignore_features=None, algorithms=None, exclude=('XCS', 'eLCS'), cv_partitions=3, exclude_plots=None, categorical_cutoff=10, sig_cutoff=0.05, scale_data=True, impute_data=True, multi_impute=True, show_plots=False, scoring_metric='balanced_accuracy', random_state=None)

Bases: Job

This ‘Job’ script conducts exploratory analysis on the new replication dataset then applies and evaluates all trained models on one or more previously unseen hold-out or replication study dataset(s). It also generates new evaluation figure. It does not deal with model feature importance estimation as this is a part of model training interpretation only. This script is run once for each replication dataset in rep_data_path.

eval_model(algorithm, cv_count, x_test, y_test)

impute_rep_data(cv_count, cv_rep_data, all_train_feature_list, cat_features, quant_features)

run()

scale_rep_data(cv_count, cv_rep_data, all_train_feature_list)