A Hybrid Spatiotemporal Deep Learning Model for Short-Term Metro Passenger Flow Prediction

Table 5

The comparison of different models (inbound passenger flow).

Model	Terminal stations			Transfer stations			Regular stations
	RMSE	MAE	MAPE (%)	RMSE	MAE	MAPE (%)	RMSE	MAE	MAPE (%)
HSTDL	32.246	18.782	31.6	52.794	25.237	16.2	18.264	12.167	22.1
GBRT	38.39	20.233	39.9	78.106	43.088	21.1	25.678	16.077	27.4
CNN	61.347	28.421	41.0	89.013	44.962	29.9	37.907	20.022	37.2
LSTM	49.951	26.667	42.6	85.329	43.066	29.7	35.913	19.995	37.5
MLP	63.998	33.195	50.0	98.637	47.86	32.5	50.338	25.531	41.2
ARIMA	77.124	51.186	61.3	125.434	56.093	36.6	60.691	36.057	47.0