Rail Transit Prediction Based on Multi-View Graph Attention Networks

Table 1

Accuracy results of rail passenger flow prediction experiment.

Methods	July		September
	MAE	RMSE	MAE	RMSE
ARIMA	18.34/20.12/23.32	29.14/33.37/36.81	19.64/21.31/24.06	30.74/34.30/36.66
SVR	14.73/16.55/18.26	25.24/31.33/32.71	15.89/16.71/17.36	28.30/33.01/34.09
LSTM	10.76/12.27/12.86	21.22/22.33/23.74	11.95/12.56/13.77	23.95/26.43/28.34
T-GCN	10.88/12.46/12.73	20.93/22.72/24.61	11.00/12.77/13.75	23.98/25.98/28.35
STGCN	9.04/10.29/10.88	19.38/20.49/22.51	10.99/11.95/13.42	21.03/23.03/24.06
DCRNN	8.41/9.73/11.56	19.43/23.76/25.77	8.72/9.24/12.73	20.94/22.01/25.82
MV-GAT	8.35/9.57/11.60	19.41/21.84/22.38	8.67/9.13/12.45	20.85/22.13/25.67