Application Of Artificial Neural Networks For Classification And Prediction Of Air Quality Classes
Free (open access)
219 - 228
J. Skrzypski, K. Kamiński, E. Jach-Szakiel & W. Kamiński
In this study, the results of investigations which enable an extension of the mathematical methods supporting air quality management in cities are presented. The actions were focused on the development of neural models of classification and prediction of the air quality classes (in respect of PM10 dust concentrations). The air quality class on a following day was predicted. The aim of modelling was to predict the air quality classes in the afternoon and in the evening when PM10 concentrations attained the daily maxima. The monitoring of PM10 concentration and the meteorological data for winter periods in 2004–2007 was used. The artificial neural network methods (ANN) with a simultaneous application of data compression methods were tested. The results of the air quality prediction are satisfactory. The accurate prognoses are predominant. The percentage of wrong prognoses is relatively small. The investigations confirm that neural prediction models allow good results to be obtained of the air quality class prediction. The results of the research prove that the tested models may be applied in the practice of air quality management in cities. Keywords: classification, prediction, air quality, big cities, PM10, artificial neural network, modelling.
classification, prediction, air quality, big cities, PM10, artificial neural network, modelling