Table 3.

All models performance in each dataset.

Dataset	Model	MSE ± STD
		Train Error	Test Error
Beijin PM2.5	LSTM-0	0.021 ± 0.020	0.016 ± 0.009
	LSTM-mean	0.021 ± 0.016	0.015 ± 0.011
	B-LSTM	0.016 ± 0.008	0.010 ± 0.004
	F-LSTM	0.180 ± 0.324	0.172 ± 0.323
	BVS-LSTM	0.013 ± 0.006	0.010 ± 0.004
	FBVS-LSTM	0.012 ± 0.004	0.011 ± 0.005
Italy Air Quality	LSTM-0	0.122 ± 0.123	0.130 ± 0.142
	LSTM-mean	0.063 ± 0.078	0.066 ± 0.082
	B-LSTM	0.027 ± 0.003	0.027 ± 0.009
	F-LSTM	0.030 ± 0.007	0.029 ± 0.015
	BVS-LSTM	0.031 ± 0.011	0.031 ± 0.013
	FBVS-LSTM	0.023 ± 0.002	0.024 ± 0.006
Beijing Multi-Site Air-Quality	LSTM-0	0.049 ± 0.029	0.06± 0.032
	LSTM-mean	0.038 ± 0.015	0.03 ± 0.023
	B-LSTM	0.031 ± 0.011	0.03 ± 0.016
	F-LSTM	0.179 ± 0.3	0.148 ± 0.25
	BVS-LSTM	0.034 ± 0.008	0.040 ± 0.024
	FBVS-LSTM	0.026 ± 0.019	0.031 ± 0.002