Table 1.

Classification metrics: accuracy (ACC), area under the curve (AUC) and f₁-score of three data combinations, two classification models (RF and GBRT), and two training/test splitting methods (KF and RS). In both KF and RS results, using combination 2 led to significantly improved accuracy (from ≤ 0.745 to ≥ 0.950), for both RF and GBRT. Combination 3 further increased the metrics. Similar pattern was observed for AUC and f₁-score: using combination 2 increased the results (AUC: from ≤ 0.813 to ≥ 0.972, f₁-score: from ≤ 0.747 to ≥ 0.950). Std. stands for standard deviation

Method		Data		Model	Mean ± std
		Training	Test		ACC	AUC	f1-score
KF	(1)	Ori	Ori	RF	0.726 ± 0.020	0.788 ± 0.007	0.726 ± 0.018
				GBRT	0.738 ± 0.009	0.797 ± 0.008	0.740 ± 0.009
	(2)	Aug	Ori	RF	0.955 ± 0.005	0.978 ± 0.005	0.955 ± 0.003
				GBRT	0.956 ± 0.004	0.982 ± 0.003	0.956 ± 0.003
	(3)	Aug	Aug	RF	0.966 ± 0.002	0.992 ± 0.001	0.966 ± 0.002
				GBRT	0.967 ± 0.002	0.993 ± 0.001	0.967 ± 0.002
RS	(1)	Ori	Ori	RF	0.692 ± 0.015	0.758 ± 0.019	0.671 ± 0.019
				GBRT	0.745 ± 0.012	0.813 ± 0.017	0.747 ± 0.012
	(2)	Aug	Ori	RF	0.950 ± 0.006	0.972 ± 0.007	0.950 ± 0.006
				GBRT	0.960 ± 0.005	0.984 ± 0.004	0.960 ± 0.004
	(3)	Aug	Aug	RF	0.950 ± 0.003	0.985 ± 0.002	0.950 ± 0.003
				GBRT	0.965 ± 0.003	0.992 ± 0.001	0.966 ± 0.003

Ori. original, Aug. augmented