AGCN-T: A Traffic Flow Prediction Model for Spatial-Temporal Network Dynamics

Table 3

Prediction performance of each method on PeMSD7.

Time (min)	Evaluation Metrics	PeMSD7
		ARIMA	FC-LSTM	T-GCN	ST-GCN	DCRNN	AGCN-T
15	MAE	9.3507	9.0169	6.4001	6.3835	6.1991	4.3097
	MAPE (%)	12.6691	12.0103	10.5345	10.3706	10.2325	10.0922
	RMSE	9.9915	9.8204	7.3981	6.8004	6.1723	4.8063
30	MAE	9.1279	9.4351	7.2408	6.2026	6.2422	4.2688
	MAPE (%)	13.3908	12.4609	11.2786	10.5237	10.3404	10.1716
	RMSE	9.7209	9.6602	7.7804	6.6337	6.6007	4.7951
45	MAE	9.3672	9.1027	7.4326	7.0218	6.4205	4.2936
	MAPE (%)	13.0813	12.9773	11.3014	11.1254	10.1624	10.0164
	RMSE	10.8952	9.6274	8.1323	7.3609	6.9523	4.7809
60	MAE	10.1145	9.2512	8.1004	7.3711	7.0213	4.4184
	MAPE (%)	13.8922	12.6821	11.2004	11.1055	11.3105	10.3243
	RMSE	10.9011	9.8308	8.5835	7.6606	8.9028	4.9245