A Neural Network Based Model for PM2.5 Air Pollutant Forecasting

Accurate forecasting of fine particulate matter concentration in cities is an important problem that can be solved with efficient methods as those provided by computational intelligence, which apply a data driven approach. An example of such method is given by artificial neural networks that are universal approximators, providing very good solutions to time series forecasting. The paper presents a neural network model for the prediction of PM2.5 air pollutant concentration, based on feed-forward neural networks and recurrent neural networks, which was developed by experiment, starting with different settings of the artificial neural network architecture. A case study of short-term PM2.5 forecasting based on the proposed neural model is detailed in the paper.