estimation of ambient air pm2.5 concentration using mlp and rbf

Background: exposure to air pollutants, such as pm2.5 is recognized as a significant health risk, contributing to the development of various diseases, and increased risk of premature mortality.methods: multilayer perceptron (mlp) and radial basis function (rbf) neural networks, were used to predict the hourly concentration of pm2.5 in isfahan, iran. the mlp model was designed with five input variables, including pm2.5 concentration and weather characteristics, ten hidden layers, and a single output layer. the dataset was divided into three subsets: 70% for training, 15% for testing, and 15% for validation.results: the results showed that the average concentration of pm2.5 was 26.5 μg/m3. the root mean square error (rmse) was estimated as 6.49 μg/m3. increasing the input data resulted in a slight reduction in network error, with the rbf model, utilizing 1450 inputs and an rmse of 6.47, achieving the same accuracy as the mlp model with 10 inputs.conclusion: given that the pm2.5 concentration estimates from the rbf and mlp models deviated by less than 23 and 25%, respectively, compared to the observed concentrations, both mlp and rbf can be regarded as reliable tools for predicting pm2.5 levels.