Air Quality Prediction Model Based on Spatiotemporal Data Analysis and Metalearning

Table 4

Comparison of mean prediction results by different methods among 36 monitoring stations in Beijing (RMSE, MAE, ACC).

Methods	+6 h			+12 h			+24 h			+48 h
	RMSE	MAE	ACC	RMSE	MAE	ACC	RMSE	MAE	ACC	RMSE	MAE	ACC
ARIMA	53.3	40.7	0.622	70.7	59.2	0.432	94.1	68.2	0.321	104.3	78.2	0.201
LSTM	48.2	34.5	0.723	57.3	47.3	0.582	71.2	55.1	0.473	85.7	66.3	0.367
ST-DNN	35.6	23.8	0.761	52.2	37.1	0.67	63.2	49.3	0.546	70.4	57.7	0.474
GAT-LSTM	28.5	20.1	0.799	47.9	34.8	0.698	55.7	45.2	0.621	65.8	48.9	0.501