Table 3.

Comparison of the performance in six traditional models and the proposed DXLR model

Models	Metric [Mean ± SD]
	Precision	Recall	Accuracy	F1 score	AUC
LR	0.538 ± 0.016	0.649 ± 0.016	0.712 ± 0.008	0.589 ± 0.013	0.766 ± 0.009
SVM	0.540 ± 0.016	0.644 ± 0.015	0.713 ± 0.008	0.587 ± 0.013	0.765 ± 0.009
DT	0.647 ± 0.015	0.827 ± 0.020	0.802 ± 0.007	0.726 ± 0.011	0.879 ± 0.006
RF	0.681 ± 0.015	0.833 ± 0.013	0.823 ± 0.007	0.749 ± 0.011	0.905 ± 0.005
XGBoost	0.714 ± 0.014	0.878 ± 0.012	0.850 ± 0.006	0.788 ± 0.010	0.928 ± 0.005
LightGBM	0.674 ± 0.015	0.836 ± 0.011	0.820 ± 0.007	0.746 ± 0.011	0.901 ± 0.005
DXLR	0.723 ± 0.014	0.892 ± 0.012	0.857 ± 0.007	0.798 ± 0.010	0.934 ± 0.004
P-valuea	< 0.0001	< 0.0001	< 0.0001	< 0.0001	< 0.0001

^a: t-test for the DXLR model and the best performing traditional model (XGBoost); The bold font is the best performing model