DCAST: A Spatiotemporal Model with DenseNet and GRU Based on Attention Mechanism

Table 3

BikeNYC and TaxiBJ results compared with the baseline models (in order to speed up the experiment, the results of reference [21] provide some baseline models).

Dataset	Model	RMSE	MAE
BikeNYC	HA	11.76	–
	ARIMA	10.07	–
	SARIMA	10.56	–
	VAR	9.92	–
	DeepST	7.43	–
	ST-ResNet	6.33	2.95
	PredCNN	5.94	2.82
	DCAST	5.53  0.10	2.64
TaxiBJ	HA	57.69	–
	ARIMA	22.78	–
	SARIMA	26.88	–
	VAR	22.88	–
	ST-ANN	19.57	–
	DeepST	18.18	–
	ST-ResNet	16.69	9.71
	PredCNN	15.90	9.22
	DCAST	15.700.10	9.13