Table 3.

Performance statistics for the RF classifier models (1–6 refer to Table 1) for the training, validation and test sets

Model #	1		2		3		4		5		6
	Valid	Test	Valid	Test	Valid	Test	Valid	Test	Valid	Test	Valid	Test
Acc	$0.942\pm 0.000$	$0.762$	$0.943\pm 0.002$	$0.937$	$0.944\pm 0.000$	$0.948$	$0.944\pm 0.000$	$0.945$	$0.946\pm 0.000$	$0.949$	$0.946\pm 0.000$	$0.949$
F1-score	$0.072\pm 0.003$	$0.071$	$0.072\pm 0.012$	$0.066$	$0.137\pm 0.010$	$0.135$	$0.143\pm 0.002$	$0.141$	$0.187\pm 0.005$	$0.196$	$0.189\pm 0.002$	$0.199$
Se	$0.208\pm 0.009$	$0.192$	$0.203\pm 0.031$	$0.211$	$0.408\pm 0.031$	$0.385$	$0.429\pm 0.006$	$0.427$	$0.573\pm 0.018$	$0.588$	$0.581\pm 0.009$	$0.596$
Sp	$0.950\pm 0.000$	$0.955$	$0.951\pm 0.002$	$0.945$	$0.950\pm 0.000$	$0.954$	$0.950\pm 0.000$	$0.951$	$0.950\pm 0.000$	$0.953$	$0.950\pm 0.000$	$0.953$
PPV	$0.044\pm 0.002$	$0.043$	$0.044\pm 0.007$	$0.039$	$0.082\pm 0.006$	$0.082$	$0.086\pm 0.001$	$0.085$	$0.111\pm 0.003$	$0.118$	$0.113\pm 0.001$	$0.119$
NPV	$0.991\pm 0.000$	$0.991$	$0.991\pm 0.000$	$0.991$	$0.993\pm 0.000$	$0.993$	$0.993\pm 0.000$	$0.994$	$0.995\pm 0.000$	$0.995$	$0.995\pm 0.000$	$0.996$
AUROC	$0.774\pm 0.011$	$0.783\mathrm{ }\left(0.773-0.790\right)$	$0.676\pm 0.028$	0.690 (0.684 − 0.706)	0.840$\pm$0.019	0.839 (0.826 − 0.843)	0.820$\pm$0.002	0.823 (0.814 − 0.833)	0.894$\pm$0.006	0.901 (0.893 − 0.906)	0.900$\pm$0.004	0.909 (0.903–0.914)

For train and validation, the $\mathrm{mean}\pm \mathrm{standard}\mathrm{ }\mathrm{deviation}$ is reported. For test AUROC, 95% CI is reported as (lower, upper) near the reported score. Best AUROC performance is obtained for model 6 (underlined).