Table 1.

Average AdaBoost model performance over the 5 test folds, and total results using the leave-one-out approach, in terms of accuracy, recall, precision, and F1-Score per class, for different sets of features (only HBR features or TBR features). The “absolute” results refer to the scores computed over the total and single confusion matrix obtained by putting together the predictions over the 5 test folds.

Class	Metrics [%]	5-Fold Cross Validation				Leave-one-out
		HBR		TBR		HBR	TBR
		$\mu +\sigma$	“absolute”	$\mu +\sigma$	“absolute”	μ (=“absolute”) + σ	μ (=“absolute”) + σ
	Accuracy	$86.7\pm 7.5$	86.7	$73.3\pm 19.0$	73.3	90.0 ± 30.5	80.0 ± 40.7
PwMS	Recall	$88.3\pm 16.2$	88.2	$71.7\pm 31.0$	70.6	94.1 ± 23.6	88.2 ± 32.3
	Precision	$91.0\pm 12.4$	88.2	$82.7\pm 16.7$	80.0	88.8 ± 31.5	78.9 ± 40.8
	F1-Score	$89.6\pm 14.3$	88.2	$76.8\pm 23.1$	75.0	91.4 ± 28.0	83.3 ± 37.3
Control	Recall	$83.3\pm 23.6$	84.6	$76.7\pm 22.4$	76.9	84.6 ± 36.1	69.2 ± 46.2
	Precision	$88.3\pm 16.2$	84.6	$74.7\pm 25.6$	66.7	91.7 ± 27.6	81.8 ± 38.6
	F1-Score	$85.8\pm 19.9$	84.6	$75.7\pm 16.5$	71.4	88.0 ± 32.5	75.0 ± 43.3