MDGCN: Multiple Graph Convolutional Network Based on the Differential Calculation for Passenger Flow Forecasting in Urban Rail Transit

Table 3

The performance in different baselines.

Model	SH
	RMSE			MAE
	10 min	15 min	30 min	10 min	15 min	30 min
HA	69.812	89.210	95.574	35.954	49.034	54.601
ARIMA	51.044	81.012	156.256	29.057	44.257	82.964
SVR	30.454	36.214	39.754	17.207	19.903	21.169
GBDT	29.935	39.364	32.542	16.561	20.001	17.342
LSTM	28.059	26.541	29.354	15.705	14.672	15.584
Conv-GCN	22.154	21.097	25.645	11.254	14.360	14.391
ResLSTM	26.387	29.548	29.642	14.087	15.265	16.005
GCN + LSTM	20.165	21.364	24.608	13.904	14.367	13.927
MDGCN	17.041	20.962	22.354	10.552	12.786	13.452
Model	XM
	RMSE			MAE
	10 min	15 min	30 min	10 min	15 min	30 min
HA	67.241	85.455	95.147	35.264	47.751	53.813
ARIMA	50.695	79.568	175.265	28.564	43.561	95.556
SVR	29.365	35.524	36.524	16.454	18.428	19.658
GBDT	33.635	37.652	50.562	17.058	19.952	26.895
LSTM	28.545	26.421	30.241	14.214	16.754	19.241
Conv-GCN	19.421	20.145	24.245	12.256	13.978	15.842
ResLSTM	24.007	23.557	26.352	14.545	15.812	18.731
GCN + LSTM	18.954	19.145	22.296	11.925	10.770	13.754
MDGCN	16.713	16.927	18.560	9.924	10.247	11.155