A Novel Methodology for Credit Spread Prediction: Depth-Gated Recurrent Neural Network with Self-Attention Mechanism

Table 3

Statistical performance indicators of the prediction models.

	DGRNN				Machine learning			VAR
	DJANET	DGRU	DLSTM	DRNN	SVR	MLP	RF
MAE	5.6770	7.7149	8.4230	11.9192	12.0704	10.1098	9.8887	12.7386
DM	—	1.8761	1.9799	2.5902	2.7406	2.4763	2.1659	3.1721
MAPE	0.0479	0.0635	0.0728	0.1046	0.2271	0.0863	0.0763	0.3826
D.M.	—	1.6973	1.9477	3.7901	5.6809	2.7217	2.0705	8.3050
RSR	0.4198	0.5549	0.5995	0.8277	1.6920	0.7263	0.7076	1.8340
SDAPE	0.0316	0.0437	0.0442	0.0570	0.1154	0.0526	0.0471	0.2008

Note:, , and indicate that D.M. statistics are significant at the 10%, 5%, and 1% levels, respectively; DJANET is the benchmark in the D.M. test; boldface represents the optimal value under different evaluation criteria; the following tables are the same.