Predictive Ability of Improved Neural Network Models to Simulate Pollutant Dispersion

Table 4

Performance of ANN models for prediction of CO, PM10, and PM2.5 (Dhaka).

Model	𝐹mean	𝑃mean	𝜎𝑜	𝜎𝑃	MBE	MSE	RMSE	𝑟^2	𝜉
ANN-MT-CO	4.8	5.37	2.03	1.95	0.73	6.9	1.98	0.47	0.77
ANN-M-CO	4.8	6.89	2.03	1.79	1.59	8.8	2.81	0.38	0.62
ANN-T-CO	4.8	8.37	2.03	0.81	3.16	18.1	5.07	0.09	0.37
ANN-MT-PM10	132.5	154.2	5.18	4.97	1.79	18.65	5.52	0.43	0.73
ANN-M-PM10	132.5	243.1	5.18	4.52	3.84	23.10	7.47	0.33	0.54
ANN-T-PM10	132.5	283.2	5.18	2.10	7.79	46.11	13.71	0.07	0.31
ANN-MT-PM2.5	97.6	118.1	4.79	4.53	1.65	16.98	5.07	0.33	0.69
ANN-M-PM2.5	97.6	191.2	4.79	4.12	3.48	21.24	6.69	0.29	0.51
ANN-T-PM2.5	97.6	219.4	4.79	1.89	7.11	42.10	12.43	0.05	0.29

F: field concentration; P: predicted concentration; CO concentration in ppm; PM10/PM2.5 concentration in 𝜇g/m³.