A Novel Prediction Model for Car Body Vibration Acceleration Based on Correlation Analysis and Neural Networks

Table 3

Performance of the FFNN-structure prediction model.

Method	Set	MSE	R	MAE	Running time (s)
BQ	Training	1.67e-04	0.21	0.0095	4.04
	Validation	1.88e-04	-0.10	0.0105
	Test	5.43e-04	0.21	0.0191
CGB	Training	5.29e-05	0.75	0.0053	2.21
	Validation	4.81e-05	0.58	0.0052
	Test	1.15e-04	0.71	0.0085
CGF	Training	1.42e-04	0.25	0.0089	1.52
	Validation	1.94e-04	0.24	0.0112
	Test	3.71e-04	0.38	0.0163
CGP	Training	1.01e-04	0.52	0.0077	1.81
	Validation	8.24e-05	0.34	0.0068
	Test	3.38e-04	0.33	0.0136
OSS	Training	5.79e-04	-0.22	0.0196	1.68
	Validation	4.04e-04	-0.44	0.0174
	Test	5.77e-04	-0.43	0.0208
RB	Training	2.91e-05	0.89	0.0040	1.77
	Validation	2.52e-05	0.80	0.0038
	Test	7.86e-05	0.71	0.0057
SCG	Training	1.32e-04	0.43	0.0088	1.09
	Validation	1.31e-04	0.53	0.0094
	Test	7.86e-05	0.30	0.0213
LM	Training	2.61e-05	0.92	0.0038	16.20
	Validation	4.91e-05	0.87	0.0053
	Test	2.21e-04	0.86	0.0134
BR	Training	1.77e-05	0.94	0.0031	289.73
	Test	2.38e-05	0.78	0.0033