Portal de Periódicos da CAPES

Prediction of Peak Concentrations of PM10 in the Area of Campo de Gibraltar (Spain) Using Classification Models

2011; Springer Science+Business Media; Linguagem: Inglês

10.1007/978-3-642-19644-7_22

1867-5670

Eva Muñoz García, M. Luz Martín Rodríguez, M.J. Jiménez–Come, Francisco Trujillo Espinosa, Ignacio Turias Domínguez,

Vehicle emissions and performance

A comparative study of different classification methods that would en able predicting the peaks of pollutant concentrations in critical meteorological sit uations is carried out because of particulates emissions that cause the widespread existing industry in the area of Campo de Gibraltar. The classification methods used in this study are k-nearest-neighbour, Bayesian classifier, Backpropagation Multilayer Neural Network, and Support Vector Machine to predict daily mean concentrations peaks. The prediction of particulate matter (PM10) concentrations was performed on the basis of their concentration lagged and using other exogenous information as: temperature, humidity, wind speed and wind direction data. In order to avoid the curse of dimensionality, Principal Component Analysis (PCA) and Fisher Linear Discriminant Analysis (Fisher LDA) were applied as feature se lection methods. The study results indicate that the support vector machine models are able to give better predictions with fewer fractions of false peaks detected than the rest of classification models.