notesum.ai

Published at November 18

Unveiling the Inflexibility of Adaptive Embedding in Traffic Forecasting

cs.AI

Released Date: November 18, 2024

Authors: Hongjun Wang, Jiyuan Chen, Lingyu Zhang, Renhe Jiang, Xuan Song

Arxiv: http://arxiv.org/abs/2411.11448v1

Refer to caption

Model	PEMS07 $\rightarrow$ PEMS04			PEMS03 $\rightarrow$ PEMS04			PEMS07 $\rightarrow$ PEMS08			PEMS03 $\rightarrow$ PEMS08
Model	MAE	RMSE	MAPE	MAE	RMSE	MAPE	MAE	RMSE	MAPE	MAE	RMSE	MAPE
DGCRN	28.52	42.86	23.77%	32.68	47.95	26.32%	21.85	32.96	15.53%	26.32	38.18	34.43 %
STID	25.32	38.32	18.85%	32.70	48.58	26.53%	19.93	29.89	13.63%	26.88	39.04	32.01%
D²STGNN	31.61	47.47	23.28%	35.01	49.57	28.43%	21.41	32.13	13.37%	31.10	42.74	35.68%
GWNet	26.05	39.07	19.02%	28.47	42.99	23.44%	18.81	28.79	12.02%	23.84	35.00	29.67%
STAEformer	25.20	37.96	18.90%	27.97	25.72	22.22%	18.44	28.53	11.79%	23.58	31.88	28.37%
Model	San Diego $\rightarrow$ Los Angeles			San Diego $\rightarrow$ Bay Area			Bay Area $\rightarrow$ San Diego			Bay Area $\rightarrow$ Los Angeles
Model	MAE	RMSE	MAPE	MAE	RMSE	MAPE	MAE	RMSE	MAPE	MAE	RMSE	MAPE
DGCRN	27.80	41.90	20.50%	27.62	42.34	21.30%	25.42	38.85	17.20%	25.98	40.05	16.40%
STID	25.82	39.56	17.57%	25.65	39.97	18.32%	23.61	36.11	15.99%	24.15	37.26	15.27%
D²STGNN	30.75	46.20	22.65%	30.55	46.70	23.55%	28.10	42.90	19.00%	28.72	44.20	18.10%
GWNet	25.61	38.64	20.20%	25.53	39.26	20.31%	23.42	35.78	18.50%	23.95	36.90	17.60%
STAEformer	24.95	37.50	17.10%	24.80	37.80	17.80%	22.80	35.00	15.50%	23.30	36.10	14.80%