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

Table 4

Prediction performance of each method on Seattle.

Time (min)	Evaluation Metrics	Seattle
		ARIMA	FC-LSTM	T-GCN	ST-GCN	DCRNN	AGCN-T
15	MAE	8.5586	7.3821	6.2605	6.1344	6.1285	4.2835
	MAPE (%)	15.3808	14.6208	13.4823	12.3895	12.0035	12.5774
	RMSE	8.8708	8.4208	6.7457	6.6926	6.2604	4.6178
30	MAE	8.4039	7.4755	6.3435	6.6204	6.4912	4.3161
	MAPE (%)	16.2407	15.5821	14.6907	14.1455	12.4211	12.0112
	RMSE	8.7053	7.8238	6.7307	6.9121	6.8134	4.7331
45	MAE	9.3711	8.1826	6.3942	6.0143	6.2233	4.2224
	MAPE (%)	16.7621	15.6608	13.3916	13.2402	13.1903	11.9529
	RMSE	9.6308	9.6313	6.9763	6.6647	6.7311	4.6066
60	MAE	9.3907	8.0623	7.0564	7.2304	7.4671	4.4373
	MAPE (%)	16.7206	15.3104	14.4216	13.1615	13.5909	12.4127
	RMSE	10.3506	9.1268	8.4527	7.9032	7.9124	4.9595