Table 1

The effect of different data processing methods on model prediction performance (bootstrapping)

	AUC		Accuracy		Precision		Recall rate		F1 value
	Mean±SD	95% CI	Mean±SD	95% CI	Mean±SD	95% CI	Mean±SD	95% CI	Mean±SD	95% CI
Data filling
No filling	0.786±0.101	0.785 to 0.787	0.770±0.070	0.769 to 0.771	0.437±0.162	0.435 to 0.438	0.546±0.208	0.544 to 0.548	0.460±0.142	0.459 to 0.461
Simple filling	0.687±0.094	0.686 to 0.688	0.761±0.076	0.760 to 0.761	0.455±0.180	0.453 to 0.456	0.491±0.165	0.489 to 0.492	0.442±0.126	0.441 to 0.443
RF filling	0.677±0.095	0.676 to 0.678	0.759±0.077	0.758 to 0.760	0.446±0.181	0.444 to 0.447	0.488±0.162	0.487 to 0.490	0.440±0.129	0.439 to 0.441
RF improve filling	0.678±0.092	0.677 to 0.678	0.756±0.077	0.755 to 0.757	0.443±0.179	0.442 to 0.445	0.485±0.161	0.483 to 0.486	0.435±0.125	0.434 to 0.436
p value	p<0.0001		p<0.0001		p<0.0001		p<0.0001		p<0.0001
Data sampling
No sampling	0.738±0.101	0.737 to 0.739	0.823±0.050	0.822 to 0.823	0.585±0.229	0.583 to 0.588	0.390±0.178	0.388 to 0.391	0.441±0.172	0.439 to 0.442
Random over sampler	0.718±0.109	0.717 to 0.719	0.765±0.070	0.764 to 0.765	0.437±0.154	0.435 to 0.438	0.531±0.189	0.529 to 0.533	0.457±0.135	0.456 to 0.458
Random under sampler	0.696±0.106	0.695 to 0.697	0.710±0.069	0.709 to 0.711	0.364±0.107	0.363 to 0.365	0.596±0.161	0.594 to 0.597	0.441±0.109	0.440 to 0.442
SMOTE over sampler	0.683±0.100	0.682 to 0.684	0.755±0.067	0.754 to 0.755	0.416±0.137	0.414 to 0.417	0.490±0.143	0.488 to 0.491	0.435±0.113	0.434 to 0.436
Borderline SMOTE	0.699±0.104	0.698 to 0.700	0.755±0.072	0.755 to 0.756	0.424±0.143	0.422 to 0.425	0.506±0.143	0.505 to 0.508	0.446±0.115	0.445 to 0.447
p value	p<0.0001		p<0.0001		p<0.0001		p<0.0001		p<0.0001
Variable selection
No selection	0.702±0.109	0.702 to 0.703	0.758±0.078	0.758 to 0.759	0.440±0.184	0.438 to 0.441	0.493±0.187	0.492 to 0.494	0.434±0.137	0.433 to 0.435
Lasso selection	0.713±0.105	0.712 to 0.713	0.761±0.074	0.760 to 0.761	0.447±0.173	0.445 to 0.448	0.513±0.177	0.512 to 0.514	0.448±0.128	0.447 to 0.449
Boruta selection	0.706±0.103	0.705 to 0.707	0.766±0.073	0.765 to 0.766	0.449±0.170	0.448 to 0.450	0.501±0.166	0.500 to 0.503	0.450±0.127	0.449 to 0.451
p value	p<0.0001		p<0.0001		p<0.0001		p<0.0001		p<0.0001

AUC, area under curve; RF, random forest; SMOTE, synthetic minority oversampling technique.