Online Traffic Accident Spatial-Temporal Post-Impact Prediction Model on Highways Based on Spiking Neural Networks

Table 9

Prediction performance comparison of spatial-temporal impact.

Method	4-lane (19%)		5-lane (27%)		6-lane (54%)		Average
	MAPE (%)	RMSE	MAPE (%)	RMSE	MAPE (%)	RMSE	MAPE (%)	RMSE
Full-recovery time (minute)
Shockwave theory	30.88	33.96	30.05	32.77	27.28	29.88	28.71	31.44
Nonlinear regression	29.07	31.54	27.96	31.08	26.03	26.03	27.13	28.44
BPNNs (base)	24.55	26.05	25.37	25.73	24.07	22.36	24.51	23.97
BPNNs (ours)	19.30	18.03	20.05	17.99	17.93	15.02	18.76	16.39
Half-recovery time (min)
Shockwave theory	36.52	22.58	32.95	20.30	33.84	20.17	34.11	20.66
Nonlinear regression	27.33	18.07	27.74	17.94	26.32	17.72	26.90	17.85
BPNNs (base)	21.80	16.33	22.93	15.02	21.01	13.22	21.68	14.30
BPNNs (ours)	18.71	11.26	17.97	10.38	18.02	9.96	18.14	10.32
Maximum accumulative queue length (mile)
Shockwave theory	38.74	0.52	38.55	0.50	42.03	0.47	40.47	0.49
Nonlinear regression	29.75	0.37	28.03	0.33	32.55	0.34	30.80	0.34
BPNNs (base)	21.74	0.31	23.09	0.31	24.11	0.27	23.38	0.29
BPNNs (ours)	18.24	0.24	18.01	0.21	21.39	0.19	19.88	0.20
Average accumulative queue length (mile)
Shockwave theory	35.52	0.43	36.04	0.36	40.54	0.35	38.37	0.37
Nonlinear regression	26.22	0.26	22.34	0.19	25.62	0.25	24.85	0.23
BPNNs (base)	18.41	0.20	20.03	0.16	22.57	0.19	21.09	0.18
BPNNs (ours)	16.28	0.19	17.26	0.12	20.09	0.13	18.60	0.14