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

Table 4

Results of the prediction models (loss function = “MAE”)

	True value	DJANET	DGRU	DLSTM	DRNN	SVR	MLP	RF	VAR
9/2017	112.89	114.51^#	109.59	118.56	115.36	108.56	116.97	117.694	121.44
10/2017	107.45	106.21	107.89^#	113.95	112.15	106.05	113.28	112.76	115.04
11/2017	104.42	105.73	105.31^#	110.77	108.05	101.11	106.83	109.22	112.30
12/2017	100.40	100.97^#	102.42	107.45	105.83	97.19	105.64	105.71	111.58
1/2018	94.71	94.03	98.69	101.68	100.32	94.72^#	101.68	100.51	106.28
2/2018	96.52	95.70^#	103.21	106.88	103.22	99.78	98.77	111.14	103.21
3/2018	109.66	104.35	111.59	116.20	111.28	110.32^#	110.56	112.88	114.01
4/2018	112.47	112.63^#	118.29	123.56	116.87	117.08	115.34	124.02	116.36
5/2018	115.95	116.33^#	110.80	113.83	110.56	113.03	117.52	111.91	122.35
6/2018	123.95	122.61	117.84	121.34	118.13	117.28	124.91^#	116.67	131.73
7/2018	122.95	124.95^#	117.51	119.97	117.96	117.32	127.17	117.46	131.63
8/2018	117.78	116.48^#	113.22	115.88	115.23	112.03	122.08	115.09	127.66
9/2018	115.94	117.47	112.53	113.81	113.92	109.91	121.54	116.97^#	125.37
10/2018	117.09	115.53	113.99	113.60	113.21	110.01	116.28^#	123.65	122.99
11/2018	132.20	130.91^#	125.84	130.23	125.76	124.81	130.04	130.63	135.14
12/2018	151.47	150.50^#	149.44	158.19	146.64	150.15	149.32	152.88	152.51
1/2019	150.14	151.70	150.50^#	159.20	148.31	148.73	147.88	149.66	154.39