Skip to main content

View full-text article in PMC

. 2020 Nov 11;22(11):e23128. doi: 10.2196/23128

Table 1.

Summary of prediction results of multiple models on the test set.

Value	Models
	Logistic regression	AdaBoost^a	GBDT^b	XGBoost^c	CatBoost
AUC^d	0.84	0.9133	0.85	0.92	0.9133
Threshold	0.3962	0.4283	0.4583	0.4478	0.5063
Youden index	0.6667	0.7333	0.6333	0.7667	0.7667
95% CI of the AUC	0.6556-1.0	0.8024-1.0	0.6997-1.0	0.8142-1.0	0.7997-1.0
SD of the AUC	0.094	0.0566	0.0784	0.054	0.058
P value of the AUC	.003	<.001	.002	<.001	<.001
Accuracy	0.76	0.76	0.76	0.84	0.84
Specificity	0.8	0.9	0.8	0.9	0.8
Sensitivity	0.7333	0.6667	0.7333	0.8	0.8667
Positive predictive value	0.8462	0.9091	0.8462	0.9231	0.8667
Negative predictive value	0.6667	0.6429	0.6667	0.75	0.8
Positive likelihood ratio	3.6667	6.6667	3.6667	8	4.3333
Negative likelihood ratio	0.3333	0.3704	0.3333	0.2222	0.1667

^aAdaBoost: adaptive boosting.

^bGBDT: gradient boosting decision tree.

^cXGBoost: eXtreme Gradient Boosting.

^dAUC: area under the receiver operating characteristic curve.