An Improved CEEMDAN-FE-TCN Model for Highway Traffic Flow Prediction

Table 6

Prediction error and training time of different models based on TCN of 717490 and 718462.

Sensor ID	Model	MAE	RMSE	GEH (Average)	R-Square	Training time (s)
717490	TCN	24.98	34.33	1.31	0.9668	96
	EMD-TCN	15.62	20.17	0.82	0.9885	450
	EMD-FE-TCN	15.33	19.96	0.80	0.9888	262
	EEMD-TCN	9.71	13.01	0.51	0.9952	467
	EEMD-FE-TCN	8.76	12.07	0.45	0.9959	241
	CEEMDAN-TCN	8.93	12.16	0.47	0.9958	413
	CEEMDAN-FE-TCN	8.12	11.31	0.42	0.9964	231
	ICEEMDAN-TCN	7.61	10.54	0.41	0.9969	364
	ICEEMDAN-FE-TCN	7.36	10.34	0.39	0.9970	194
Sensor ID	Model	MAE	RMSE	GEH (Average)	R-square	Training time (s)
718462	TCN	5.31	7.01	1.00	0.8203	71
	EMD-TCN	3.64	4.73	0.69	0.9180	455
	EMD-FE-TCN	3.63	4.72	0.69	0.9186	298
	EEMD-TCN	2.52	3.40	0.48	0.9577	464
	EEMD-FE-TCN	2.49	3.38	0.47	0.9583	262
	CEEMDAN-TCN	2.31	3.13	0.44	0.9642	446
	CEEMDAN-FE-TCN	2.28	3.10	0.43	0.9649	236
	ICEEMDAN-TCN	2.22	3.05	0.42	0.9660	380
	ICEEMDAN-FE-TCN	2.10	2.96	0.39	0.9681	205

Note.indicates the best results; ICEEMDAN means improved CEEMDAN.