Table 3.

Feature combinations and their predictive accuracy with different machine learning methods.

Group		Model	Feature combination	Accuracy	Balanced accuracy	AUC (95% CI)	Sensitivity (95% CI)	Specificity(95% CI)	Odds ratio(95% CI)
1	H vs. M-S	Neural Network	Tf+Pg+Pi+Fn+Pa+Cr+Td	0.93	0.91	0.96(0.95–0.98)	0.95(0.92–0.98)	0.87(0.78–0.95)	127.2(67.08–241.01)
	H vs. M-S	Random Forest	Tf+Pg+Fn+Td+Ec+Cr	0.93	0.89	0.96(0.95–0.97)	0.96(0.92–0.99)	0.83(0.75–0.91)	117.2(61.56–223.04)
	H vs. M-S	Support Vector Machine	Tf+Pg+Pi+Pa+Td	0.92	0.86	0.96(0.94–0.99)	0.97(0.95–1.00)	0.74(0.65–0.83)	92(48.03–176.33)
	H vs. M-S	Regularized Logistic Regression	Tf+Pg+Pi+Cr	0.92	0.88	0.97(0.95–0.98)	0.97(0.95–0.99)	0.78(0.66–0.91)	114.6(59.16–222.12)
			Average	0.93	0.88	0.96	0.96	0.81	112.75
2	H vs. Sli-M-S	Neural Network	Tf+Ec+Pg+Pa+Td	0.90	0.86	0.94(0.91–0.97)	0.93(0.90–0.97)	0.79(0.68–0.89)	50.0(29.71–84.07)
	H vs. Sli-M-S	Random Forest	Tf+Ec+Aa+Pg+Pa+Cr	0.91	0.84	0.94(0.92–0.96)	0.95(0.91–1.00)	0.72(0.65–0.80)	48.9(28.71–83.14)
	H vs. Sli-M-S	Support Vector Machine	Tf+Cr+Pa+Pi+Fn	0.89	0.79	0.94(0.91–0.96)	0.97(0.93–1.00)	0.61(0.53–0.70)	50.6(27.76–92.12)
	H vs. Sli-M-S	Regularized Logistic Regression	Tf+Cr+Pg+Pa+Aa+Fn+Pi	0.90	0.80	0.94(0.91–0.96)	0.96(0.93–1.00)	0.64(0.59–0.69)	42.7(24.73–73.62)
			Average	0.90	0.82	0.94	0.95	0.69	48.05
3	H vs. Sli	Neural Network	Tf	0.80	0.77	0.82(0.74–0.89)	0.67(0.55–0.80)	0.88(0.82–0.93)	14.89(7.67–28.91)
	H vs. Sli	Random Forest	Tf+Pg	0.78	0.77	0.81(0.75–0.88)	0.71(0.57–0.84)	0.83(0.76–0.89)	12.00(6.42–22.26)
	H vs. Sli	Support Vector Machine	Tf+Td	0.78	0.77	0.83(0.78–0.88)	0.72(0.55–0.89)	0.82(0.77–0.86)	11.70(6.33–21.68)
	H vs. Sli	Regularized Logistic Regression	Tf+Aa+Td	0.79	0.78	0.82(0.77–0.87)	0.75(0.58–0.91)	0.81(0.73–0.90)	12.80(6.85–23.87)
			Average	0.78	0.77	0.82	0.71	0.84	12.85

AUC, area under the curve; CI, confidence interval.