Table 3.

Mean test loss and area under the curve of 5-fold datasets according to the training set modification and epoch size (batch size = 16).

	Epoch size	Strategy 1 Original	Strategy 2 Class weight (1:0)	Strategy 3  × 10 data (positive only)	Strategy 4  × 10 data (all)	Overall
	50	0.2858	0.4646	0.3427	0.2978	0.3477
	100	0.2875	0.3940	0.3536	0.3485	0.3459
Loss	150	0.2894	0.3542	0.3646	0.3471	0.3388
	200	0.3249	0.4323	0.3113	0.3540	0.3556
	Overall	0.2969	0.4113	0.3431	0.3369	0.3470
	50	0.3400	0.4400	0.5200	0.3050	0.4013
	100	0.2775	0.4975	0.4825	0.2925	0.3875
Sensitivity	150	0.2875	0.3900	0.5600	0.2400	0.3694
	200	0.2825	0.5425	0.4425	0.3300	0.3994
	Overall	0.2969	0.4675	0.5013	0.2919	0.3894
	50	0.9546	0.8537	0.8933	0.9669	0.9171
	100	0.9737	0.8765	0.8926	0.9796	0.9306
Specificity	150	0.9785	0.9257	0.8873	0.9860	0.9444
	200	0.9678	0.8540	0.9344	0.9619	0.9295
	Overall	0.9686	0.8775	0.9019	0.9736	0.9304
	50	0.7941	0.7877	0.8129	0.8209	0.8039
	100	0.7951	0.7959	0.8162	0.8064	0.8034
AUC	150	0.7952	0.7945	0.8156	0.8053	0.8027
	200	0.7889	0.7927	0.8114	0.7998	0.7982
	Overall	0.7933	0.7927	0.8140	0.8081	0.8020

AUC, area under the curve.