Multigraph Aggregation Spatiotemporal Graph Convolution Network for Ride-Hailing Pick-Up Region Prediction

Table 2

Comparison if prediction performances in different models.

Dataset	Methods		Weekday		Weekend
			RMSE	MAE	RMSE	MAE
Chengdu	Traditional models	HA	53.25	32.18	54.67	33.58
		ARIMA	41.72	29.86	43.32	30.67
	Deep models	LSTM	40.81	26.35	42.73	28.49
		GRU	36.56	25.22	37.55	26.62
		MGCN	32.39	23.21	34.68	24.94
		STGCN	29.82	21.24	30.99	22.57
	Our model	MAST-GCN	20.57	13.29	21.82	14.12
Wuhan	Traditional models	HA	55.49	33.55	56.72	34.25
		ARIMA	43.89	31.87	45.76	32.36
	Deep models	LSTM	42.13	27.57	45.18	29.77
		GRU	37.52	26.47	38.37	28.17
		MGCN	33.89	24.31	34.94	25.39
		STGCN	30.76	21.52	31.8	23.71
	Our model	MAST-GCN	22.49	14.68	23.68	15.31