Predictive Ability of Improved Neural Network Models to Simulate Pollutant Dispersion

Table 3

Performance statistics of ANN models on the validation data set (Dhaka).

Model	Number of epochs	𝜂	𝜆	RMSE	𝜉
ANN-MT-CO (18 : 4 : 1)	1800	0.001	0.6	1.89	0.69
ANN-M-CO (12 : 4 : 1)	1300	0.001	0.4	2.18	0.63
ANN-T-CO (6 : 4 : 1)	300	0.001	0.7	3.39	0.40
ANN-MT-PM10 (18 : 4 : 1)	1700	0.001	0.7	2.08	0.73
ANN-M-PM10 (12 : 4 : 1)	1400	0.001	0.3	2.41	0.54
ANN-T-PM10 (6 : 4 : 1)	300	0.001	0.6	3.09	0.39
ANN-MT-PM2.5 (18 : 4 : 1)	1600	0.001	0.6	2.46	0.67
ANN-M-PM2.5 (12 : 4 : 1)	1300	0.001	0.4	2.41	0.51
ANN-T-PM2.5 (6 : 4 : 1)	300	0.001	0.6	2.89	0.40