Modelling ground-level ozone concentration using ensemble learning algorithms

Environmental risks caused by exposure to ground level ozone have significantly increased during recent years. One main producer of ozone is the photochemical reaction between volatile organic components and the anthropogenic nitrogen oxides created by vehicular traffic. Therefore the measurement and monitoring of atmospheric ozone concentration levels is important. In this paper we propose a study of the use of state-of-the-art machine learning approaches in modelling the concentration of ground level ozone. The prediction is based on concentrations of seven gases (NO2, SO2, and BTX (Benzene, Toluene, o-,m-,p-Xylene) and six meteorological parameters (ambient temperature, air pressure, wind speed, wind direction, global radiation, and relative humidity). The analysis of the results indicates that accurate models for the concentration of ground level ozone can be derived with the best performance accuracies indicated by the Ensemble Learning Algorithms. The investigation carried out compares the use of different machine learning classifiers and show that the Ensembleclassifier Bagging performs superior to standard single classifiers, such as Artificial Neural Networks and Support Vector Machines, popularly used in literature. In addition, we study the performance of the meta-classifier Bagging when different base classifiers are used in optimised configurations and compare the results thus obtained. The research conducted bridges an existing research gap in big-data analytics related to environment pollution prediction, where present research is largely limited to using standard learning algorithms such as Neural Networks and Support Vector Machines often available within popular commercial software packages.