TABLE 4.

Performance Comparison of Predictive Models: Responders

Algorithms	Clusters	AUC	Sensitivity	Specificity	PPV	NPV	Accuracy
Logistic regression
All features	1	0.757 (0.213)	0.617 (0.193)	0.800 (0.133)	0.658 (0.202)	0.790 (0.095)	0.734 (0.110)
	2	0.608 (0.297)	0.592 (0.382)	0.642 (0.171)	0.517 (0.285)	0.725 (0.224)	0.621 (0.171)
	Combination	0.682 (0.262)	0.604 (0.294)	0.721 (0.169)	0.587 (0.251)	0.757 (0.170)	0.677 (0.151)
	Whole cohort	0.638 (0.144)	0.400 (0.154)	0.760 (0.177)	0.570 (0.209)	0.642 (0.110)	0.615 (0.131)
10 features	1	0.757 (0.232)	0.500 (0.360)	0.900 (0.141)	0.658 (0.390)	0.783 (0.125)	0.759 (0.110)
	2	0.708 (0.233)	0.633 (0.343)	0.617 (0.172)	0.550 (0.270)	0.710 (0.236)	0.620 (0.207)
	Combination	0.732 (0.227)	0.566 (0.349)	0.758 (0.211)	0.604 (0.331)	0.746 (0.187)	0.689 (0.176)
	Whole cohort	0.653 (0.146)	0.371 (0.276)	0.778 (0.128)	0.422 (0.243)	0.654 (0.085)	0.609 (0.076)
Random forest
All features	1	0.710 (0.233)	0.367 (0.331)	0.900 (0.141)	0.583 (0.466)	0.723 (0.131)	0.707 (0.171)
	2	0.550 (0.220)	0.433 (0.288)	0.750 (0.204)	0.517 (0.309)	0.620 (0.167)	0.600 (0.197)
	Combination	0.630 (0.235)	0.400 (0.303)	0.825 (0.187)	0.550 (0.386)	0.671 (0.155)	0.683 (0.181)
	Whole cohort	0.706 (0.192)	0.431 (0.267)	0.794 (0.139)	0.571 (0.230)	0.677 (0.126)	0.648 (0.142)
10 features	1	0.550 (0.291)	0.233 (0.274)	0.760 (0.310)	0.370 (0.462)	0.607 (0.190)	0.570 (0.244)
	2	0.500 (0.297)	0.183 (0.254)	0.750 (0.264)	0.333 (0.471)	0.515 (0.150)	0.487 (0.217)
	Combination	0.525 (0.287)	0.208 (0.258)	0.755 (0.280)	0.351 (0.454)	0.561 (0.173)	0.528 (0.228)
	Whole cohort	0.697 (0.098)	0.443 (0.249)	0.796 (0.142)	0.548 (0.214)	0.686 (0.062)	0.655 (0.062)
XGBoost
All features	1	0.657 (0.185)	0.550 (0.273)	0.760 (0.207)	0.607 (0.330)	0.746 (0.137)	0.684 (0.118)
	2	0.477 (0.272)	0.433 (0.235)	0.550 (0.307)	0.492 (0.287)	0.540 (0.220)	0.407 (0.195)
	Combination	0.567 (0.244)	0.491 (0.255)	0.655 (0.276)	0.549 (0.306)	0.643 (0.207)	0.545 (0.211)
	Whole cohort	0.655 (0.150)	0.476 (0.223)	0.662 (0.139)	0.488 (0.152)	0.650 (0.138)	0.583 (0.136)
10 features	1	0.607 (0.343)	0.433 (0.316)	0.760 (0.207)	0.525 (0.389)	0.704 (0.148)	0.643 (0.201)
	2	0.423 (0.230)	0.392 (0.125)	0.550 (0.258)	0.457 (0.216)	0.497 (0.134)	0.475 (0.145)
	Combination	0.515 (0.299)	0.412 (0.234)	0.655 (0.251)	0.491 (0.308)	0.600 (0.173)	0.559 (0.191)
	Whole cohort	0.652 (0.065)	0.486 (0.206)	0.708 (0.106)	0.530 (0.049)	0.675 (0.086	0.616 (0.058)

AUC, area under the curve; NPV, negative predictive value; PPV, positive predictive value.