Table 4.

Performance of different models using BGE-M3 vectors on the test set of training cohort.

Model	Sensitivity (mean ± std)	Specificity (mean ± std)	PPV (mean ± std)	NPV (mean ± std)	F1-score (mean ± std)	AUC (mean ± std)
FNN	0.7895 ± 0.0455	0.9972 ± 0.0101	0.8864 ± 0.0217	0.9943 ± 0.0013	0.8351 ± 0.0199	0.9822 ± 0.0104
LightGBM	0.8603 ± 0.0104	0.9794 ± 0.0011	0.5339 ± 0.0018	0.9961 ± 0.0003	0.6589 ± 0.0021	0.9843 ± 0.0043
LR	0.9130 ± 0.0135	0.9101 ± 0.0033	0.2176 ± 0.0024	0.9974 ± 0.0004	0.3514 ± 0.0034	0.9716 ± 0.0025
RF	0.8603 ± 0.0219	0.9296 ± 0.0028	0.2507 ± 0.0042	0.9959 ± 0.0007	0.3883 ± 0.0069	0.9600 ± 0.0090
SVM	0.8927 ± 0.0113	0.9450 ± 0.0024	0.3077 ± 0.0026	0.9969 ± 0.0004	0.4577 ± 0.0040	0.9786 ± 0.0014
XGBoost	0.9251 ± 0.0047	0.9413 ± 0.0018	0.3015 ± 0.0074	0.9978 ± 0.0002	0.4547 ± 0.0079	0.9847 ± 0.0038