Table 2.

The F-1 scores delivered by the alternative prediction models with different feature sets.

Model	Feature set	Sensitivity 85% Mean (95% confidence interval)	Sensitivity 90% Mean (95% confidence interval)	Sensitivity 95% Mean (95% confidence interval)
DT	6 features (by the proposed DT-based method)	0.446 (0.446–0.447)	0.436 (0.436–0.436)	0.410 (0.409–0.410)
	6 features (by mRMRe)	0.438 (0.438–0.438)	0.428 (0.428–0.428)	0.409 (0.409–0.409)
	6 features (by LASSO)	0.437 (0.437–0.437)	0.428 (0.427–0.428)	0.400 (0.400–0.400)
	18 features	0.442 (0.442–0.442)	0.437 (0.437–0.437)	0.413 (0.412–0.413)
LR	6 features (by the proposed DT-based method)	0.440 (0.440–0.441)	0.430 (0.430–0.430)	0.399 (0.399–0.400)
	6 features (by mRMRe)	0.435 (0.435–0.435)	0.423 (0.423–0.424)	0.398 (0.398–0.399)
	6 features (by LASSO)	0.433 (0.432–0.433)	0.421 (0.421–0.421)	0.397 (0.397–0.397)
	18 features	0.446 (0.446–0.446)	0.434 (0.434–0.434)	0.406 (0.406–0.406)
DNN	6 features (by the proposed DT-based method)	0.447 (0.447–0.447)	0.437 (0.437–0.437)	0.409 (0.408–0.409)
	6 features (by mRMRe)	0.438 (0.438–0.438)	0.431 (0.430–0.431)	0.410 (0.410–0.410)
	6 features (by LASSO)	0.437 (0.437–0.437)	0.428 (0.427–0.428)	0.400 (0.400–0.400)
	18 features	0.452 (0.451–0.452)	0.444 (0.443–0.444)	0.422 (0.421–0.422)

Please refer to Supplementary Table S6 for the definitions of the F1 score.