Applying Computational Intelligence To The Classification Of Pollution Events

This paper compares three computational intelligence techniques applied to the discrimination of environmental situations associated to low air-quality events regarding the concentration of particulate matter with diameter lower than 10 micrometers. The techniques revised in this work are: Naive Bayesian Classification, Support Vector Machines and Fuzzy systems. A database extracted from the air-quality surveillance network at Bogota (Colombia) is used to train these classifiers. Results show that the support vector machine outperformed the other techniques in terms of exactitude and sensitivity. Although the fuzzy classifier and the Naive Bayes classifier did not achieve the best performances, these techniques offer interpretability about the classification problem.