Prediction of critical air quality events using support vector machines and particle swarm optimization

Abstract

In recent years several investigation prove the effects of air pollutants over human health, in this regard is necessary develop systems that allow people reduce the exposure to unhealthy air quality conditions. In This paper we propose a methodology to predict Critical Air Quality Events (CAQE) in Aburrá Valley based on Support Vector Machines (SVM) optimized with Particle Swarm Optimization (PSO) and a characterization scheme to asses the current and past tendencies of pollutants and weather behavior, analyzing the statistical behavior at different time intervals. We use a three stage methodology consisting in prepossessing, characterizations and CAQE prediction. The proposed method shows the better result for ozone CAQE prediction with an error of 30%. Due to low sensitivity among the pollutants is necessary use another machine learning technique that warranty a robust behavior working with unbalanced data.