Neural Network Forecast for Daily Average PM10 Concentrations

Abstract

Over the past years, the health impact of particulate matter (PM)

has become a very current subject. In the environmental sciences a lot

of research effort goes towards the understanding of the particulate matter

phenomenon and the ability to forecast particulate matter concentrations.

The aim of the present work is to evaluate the potential of various developed

artificial neural network models to provide reliable predictions of

PM₁₀ concentrations in Kayseri. Model structure obtained from air quality

monitoring network system performed by the Ministry of Environment

and Forestry was developed for 18 month data of Kayseri and the structure

was refined by Levenberg-Marquardt algorithm. Performance results

of artificial neural network models was compared with multiple regression

analysis and neural network give better predictions than multivariate

regression models.