TABLE 5.

5-fold performance comparison of different feature selection algorithms in the deep learning-based prediction model.

	Mi			t-SNE			The proposed model
	Cost	Accuracy	AUROC	Cost	Accuracy	AUROC	Cost	Accuracy	AUROC
1	1.9900	0.6156	0.4309	0.6508	0.7143	0.5000	0.0004	0.9999	0.9998
2	3.3951	0.6230	0.5000	0.6574	0.6230	0.5000	0.0171	0.9959	0.9967
3	6.7629	0.4298	0.5000	0.8742	0.4551	0.5077	0.1822	0.9849	0.9808
4	1.4862	0.5076	0.5028	0.9059	0.4702	0.4677	0.1213	0.9764	0.9772
5	3.2534	0.6534	0.5000	0.5965	0.6754	0.6193	0.0454	0.9945	0.9958
Average	3.3775	0.5659	0.4867	0.7370	0.5876	0.5189	0.0732	0.9903	0.9916

The average Cost, Accuracy and AUROC of DEG+DMP feature selection algorithm are used in the prediction model based on deep learning.