Table 10.

Comparison of different machine learning methods with selected features on Site B dataset.

Methods	AUC			Sensitivity (PPV = 70%)
	Mean	95% CI	p value	Mean	95% CI	p value
Random Forests	0.844	(0.833, 0.855)	–	92.6%	(90.0%, 95.1%)	–
SVM	0.852	(0.838, 0.866)	p = 0.148	95.4%	(93.2%, 97.5%)	$p=0.020$
Logistic Regression	0.798	(0.783, 0.812)	$p=0.003$	90.2%	(88.6%, 91.9%)	p = 0.110
DNN (small)	0.831	(0.805, 0.858)	p = 0.187	86.9%	(83.5%, 90.3%)	$p=0.003$
DNN w/ all features	0.704	(0.670, 0.738)	$p<0.001$	75.38%	(71.6%, 79.2%)	$p=0.001$
WD Net w/ all features	0.769	(0.753, 0.786)	$p<0.001$	85.6%	(82.6%, 88.7%)	$p=0.004$