The Prediction of Multistep Traffic Flow Based on AST-GCN-LSTM

Table 1

Performance comparison of different methods.

T (min)	Metrics	HA	ARIMA	SVR	DCRNN	GCN-LSTM	AST-GCN-LSTM
15	RMSE	4.2951	7.2406	4.1455	4.5000	4.1193	4.0234
	MAE	2.7815	4.9824	2.6233	3.1700	2.7701	2.7030
	Accuracy	0.7008	0.4463	0.7112	0.2913	0.7129	0.7223
30	RMSE	4.2951	6.7899	4.1628	4.5600	4.1207	4.0508
	MAE	2.7815	4.6765	2.6875	3.2300	2.7739	2.7244
	Accuracy	0.7008	0.3845	0.7100	0.2970	0.7126	0.7196
45	RMSE	4.2951	6.7852	4.1885	4.6000	4.1252	4.0587
	MAE	2.7815	4.6734	2.7359	3.2700	2.7753	2.7346
	Accuracy	0.7008	0.3847	0.7082	0.3021	0.7123	0.7172
60	RMSE	4.2951	6.7708	4.2156	4.6400	4.1262	4.0689
	MAE	2.7815	4.6655	2.7751	3.3100	2.7811	2.7403
	Accuracy	0.7008	0.3851	0.7063	0.3069	0.7119	0.7165