Table 2.

Performance of the final model after applying the risk stratification method.

Data processing	Accuracy	Precision	Sensitivity	Specificity	F1-score	AUROC	AUPRC
Internal validation
Boruta + GSCV	0.950	0.500	0.400	0.979	0.444	0.942	0.443
Boruta + GSCV + Stratified	0.960	0.720	0.474	0.989	0.571	0.971	0.741
External validation
Boruta + GSCV	0.953	0.405	0.714	0.962	0.517	0.932	0.699
Boruta + GSCV + Stratified	0.985	0.875	0.667	0.997	0.757	0.929	0.699

AUROC, the area under the receiver operating characteristic curve; AUPRC, area under the precision-recall curve; GSCV, grid search with cross-validation; RF, random forest.