TABLE 5.

Eight algorithms’ model performance in the test cohort.

Model	AUROC	SEN	SPE	AUPR	Precision	Recall	F1
RF	0.805 (0.705, 0.906)	0.760	0.773	0.550	0.528	0.760	0.623
XGBoost	0.754 (0.645, 0.863)	0.760	0.720	0.323	0.475	0.760	0.585
DT	0.650 (0.525, 0.776)	0.480	0.853	0.150	0.522	0.480	0.500
GBDT	0.832 (0.744, 0.920)	0.720	0.853	0.557	0.621	0.720	0.667
LightGBM	0.750 (0.635, 0.864)	0.840	0.640	0.485	0.438	0.840	0.575
AdaBoost	0.782 (0.678, 0.886)	0.640	0.867	0.538	0.615	0.640	0.627
CatBoost	0.817 (0.725, 0.909)	0.720	0.813	0.462	0.563	0.720	0.632
Ensemble learning model	0.797 (0.694, 0.899)	0.720	0.840	0.537	0.600	0.720	0.655