Table 2.

Different criteria of classifiers with different feature selection methods

Criterion	Model	DFS	RFE	Fisher	Nonea
ACC	SVM	76.9 ± 4.3%	66.1 ± 6.7%	65.6 ± 5.9%	66.1 ± 6.8%
	GP	78.1 ± 2.3%	63.7 ± 5.5%	63.5 ± 5.0%	55.5 ± 5.4%
	Logistic	76.7 ± 2.2%	68.9 ± 6.4%	67.4 ± 5.3%	67.7 ± 5.2%
	Adaboost	71.7 ± 2.9%	61.9 ± 3.1%	63.4 ± 4.3%	59.1 ± 3.7%
	GCN	79.5 ± 3.3%	71.1 ± 4.2%	69.6 ± 3.8%	69.8 ± 4.3%
	MLP	78.1 ± 4.7%	69.5 ± 5.6%	68.3 ± 4.8%	66.4 ± 5.3%
AUC	SVM	0.830 ± 0.048	0.700 ± 0.072	0.688 ± 0.084	0.673 ± 0.072
	GP	0.852 ± 0.036	0.697 ± 0.063	0.674 ± 0.051	0.653 ± 0.065
	Logistic	0.842 ± 0.039	0.725 ± 0.072	0.716 ± 0.076	0.687 ± 0.076
	Adaboost	0.761 ± 0.045	0.641 ± 0.037	0.672 ± 0.039	0.654 ± 0.064
	GCN	0.848 ± 0.027	0.733 ± 0.035	0.718 ± 0.030	0.724 ± 0.078
	MLP	0.851 ± 0.031	0.728 ± 0.042	0.702 ± 0.029	0.667 ± 0.076
SEN	SVM	74.0 ± 8.3%	59.9 ± 12.6%	57.4 ± 14.4%	59.1 ± 12.8%
	GP	75.7 ± 7.7%	57.3 ± 10.2%	53.8 ± 7.2%	53.2 ± 8.5%
	Logistic	74.0 ± 7.1%	62.1 ± 10.7%	60.6 ± 11.5%	60.1 ± 11.8%
	Adaboost	67.0 ± 7.7%	56.1 ± 10.1%	56.3 ± 9.1%	56.8 ± 8.7%
	GCN	78.3 ± 3.5%	68.8 ± 5.3%	65.7 ± 4.5%	66.8 ± 7.8%
	MLP	77.2 ± 4.9%	66.3 ± 5.8%	64.3 ± 6.7%	63.2 ± 8.2%
SPE	SVM	78.5 ± 6.3%	70.6 ± 8.1%	71.2 ± 11.1%	70.5 ± 9.3%
	GP	79.0 ± 7.2%	68.7 ± 7.7%	73.2 ± 3.7%	57.9 ± 9.2%
	Logistic	79.2 ± 7.5%	72.6 ± 8.2%	72.7 ± 8.3%	72.4 ± 8.8%
	Adaboost	73.3 ± 6.8%	63.0 ± 7.8%	63.3 ± 6.1%	66.9 ± 6.8%
	GCN	81.2 ± 3.6%	73.5 ± 4.9%	69.9 ± 4.3%	71.8 ± 4.8%
	MLP	79.8 ± 4.8%	71.5 ± 6.3%	69.2 ± 6.4%	68.2 ± 7.2%

^aNo feature selection adopted