Output files

After training, two .tsv files will be generated. One file with name ending with _eval.tsv has the following columns (for classification task):

Column name	Description
fs	Feature selection method (proms or proms_mo or pca_ex)
type	multiview (mo) or single view (so)
k	number of selected markers
estimator	one of the supported estimators used
repeat	repeat index
val_score	score for the validation set (comma separated)
val_pred_label	predicted labels for the validation set (comma separated)
val_label	true labels for the validation set (comma separated)
val_acc	accuracy for the validation set
val_auroc	AUROC for the validation set

The name of the other .tsv file ends with _full.tsv. This file contains the information about the final model and its performance on independent test data set (if available). It contains the following columns (for classification task):

Column name	Description
fs	Feature selection method (proms or proms_mo or pca_ex)
type	multiview (mo) or single view (so)
k	number of selected markers
estimator	the estimator used
features	name of selected markers (comma separated)
membership	a json string describe the membership of each cluster where keys are the final selected markers
mean_val_acc	mean validation accuracy for the selected markers/k/estimator combination among all evalutaion repeats
mean_val_auroc	mean validation auroc for the selected markers/k/estimator combination among all evalutaion repeats
test_score	score for the test data set (if test data set is provided)
test_pred_label	predicted labels for the test data set (if test data set is provided)
test_label	true labels for the test data set (if test data labels are also provided)
test_accuracy	accuracy for the test data set (if test data labels are also provided)
test_auroc	auroc for the test data set (if test data labels are also provided)

The final full model is generated as full_model/full_model.pkl. This can be used for making predictions on new data set.