Multicomponent Spatial-Temporal Graph Attention Convolution Networks for Traffic Prediction with Spatially Sparse Data

Table 2

Performance comparison of different methods on METR-LA with different number of nodes.

Model	24 nodes		56 nodes		80 nodes		136 nodes		207 nodes
	MAE	RMSE	MAE	RMSE	MAE	RMSE	MAE	RMSE	MAE	RMSE
HA	6.82	11.26	6.47	10.91	6.91	11.34	7.37	11.83	7.50	11.93
VAR	4.58	8.51	4.61	8.44	4.73	8.40	4.74	8.34	4.68	8.37
DCRNN	3.64	7.46	3.32	6.83	4.20	9.99	3.10	6.29	3.17	6.47
STGCN	6.06	9.35	6.03	9.11	6.03	9.12	6.09	9.08	3.65	7.46
ST-MetaNet	3.25	6.82	3.03	6.33	3.10	6.37	3.00	6.20	3.06	6.23
GraphWaveNet	3.22	6.52	3.01	6.14	3.09	6.14	3.02	6.07	3.04	6.09
MSTGACN (ours)	3.22	6.47	3.01	6.09	3.11	6.18	3.06	6.10	3.14	6.16