Prediction of PM10 and SO 2 exceedances to control air pollution in the Bay of Algeciras, Spain

摘要: In this paper, the authors apply different classification techniques in order to provide 24 h advance forecasts of daily peaks SO2 and PM10 concentrations Bay Algeciras. K-nearest-neighbours, multilayer neural network with backpropagation support vector machines (SVMs) are methods used. The aim research is obtain a suitable prediction model that would enable us predict pollutant critical meteorological situations caused by widespread existing industry population area. A resampling strategy twofold crossvalidation has been applied, using quality indexes evaluate performance models. SVM models achieved better true positive rate accuracy (ACC) indexes. Results ACC index value 0.795 for 0.755 showed ability non-peaks correctly.