streamline.runners.feature_runner module

class streamline.runners.feature_runner.FeatureImportanceRunner(output_path, experiment_name, class_label='Class', instance_label=None, instance_subset=None, algorithms=('MI', 'MS'), use_turf=True, turf_pct=True, random_state=None, n_jobs=None, run_cluster=False, queue='defq', reserved_memory=4)

Bases: object

Runner Class for running feature importance jobs for cross-validation splits.

Parameters:

• output_path –

• experiment_name –

• class_label –

• instance_label –

• instance_subset –

• algorithms –

• use_turf –

• turf_pct –

• random_state –

• n_jobs –

Returns: None

get_cluster_params(cv_train_path, experiment_path, algorithm)

run(run_parallel=False)

save_metadata()

submit_lsf_cluster_job(cv_train_path, experiment_path, algorithm)

submit_slurm_cluster_job(cv_train_path, experiment_path, algorithm)

class streamline.runners.feature_runner.FeatureSelectionRunner(output_path, experiment_name, algorithms, class_label='Class', instance_label=None, max_features_to_keep=2000, filter_poor_features=True, top_features=40, export_scores=True, overwrite_cv=True, random_state=None, n_jobs=None, run_cluster=False, queue='defq', reserved_memory=4, show_plots=False)

Bases: object

Runner Class for running feature selection jobs for cross-validation splits.

Parameters:

• output_path – path other the output folder

• experiment_name – name for the current experiment

• algorithms – feature selection algorithms from last phase

• max_features_to_keep – max features to keep (only applies if filter_poor_features is True), default=2000

• filter_poor_features – filter out the worst performing features prior to modeling,default=’True’

• top_features – number of top features to illustrate in figures, default=40)

• export_scores – export figure summarizing average fi scores over cv partitions, default=’True’

• overwrite_cv – overwrites working cv datasets with new feature subset datasets,default=”True”

• random_state – random seed for reproducibility

• n_jobs – n_jobs param for multiprocessing

Returns: None

get_cluster_params(full_path, n_datasets)

run(run_parallel=False)

save_metadata()

submit_lsf_cluster_job(full_path, n_datasets)

submit_slurm_cluster_job(full_path, n_datasets)