Skip to main content

View full-text article in PMC

. 2021 May 6;9(5):e22591. doi: 10.2196/22591

Table 4.

Performance given different feature sets.

Feature	Model	Accuracy	AUROC^a	Sensitivity	Specificity	Precision	F1
All features	DNN^b	0.9357	0.9699	0.9452	0.9253	0.9393	0.9323
Lifestyle	Random forests	0.8428	0.9195	0.9253	0.7671	0.9180	0.8358
Env^c	Decision trees	0.8000	0.8185	0.8805	0.7260	0.8688	0.7910
Env+Lifestyle	Random forests	0.8357	0.9256	0.8805	0.7945	0.8787	0.8345
Clinical questionnaire	AdaBoost^d	0.6956	0.6825	0.6666	0.7142	0.7692	0.7407

^aAUROC: area under the receiver operating characteristic curve.

^bDNN: deep neural network.

^cEnv: environmental.

^dAdaBoost: adaptive boosting.