A Novel Type-2 Adaptive Neuro Fuzzy Inference System Classifier for Modelling Uncertainty in Prediction of Air Pollution Disaster (RESEARCH NOTE)

Abstract

Type-2 fuzzy set theory is one of the most powerful tools for dealing with the uncertainty and imperfection in dynamic and complex environments. The applications of type-2 fuzzy sets and soft computing methods are rapidly emerging in the ecological fields such as air pollution and weather prediction. The air pollution problem is a major public health problem in many cities of the world. Prediction of natural phenomena always suffers from uncertainty in the environment and incompleteness of data. However, various studies have been reported for prediction of the air quality index but all of them suffer from uncertainty and imprecision associated to the incompleteness of knowledge and imprecise input measures. This article takes advantages of learning of adaptive neural networks alongside in new environment. Furthermore, it presents an Adaptive Neuro-Type-2 Fuzzy Inference System (ANT2FIS) to address the uncertainty and imprecision in air quality prediction. The data set of this study was collected from Tehran municipality official website for the last five years (2012-2017). The results reveal that the ANT2FIS prediction method is more reliable and is capable of handling uncertainty compared to the other counterpart methods. The performance results on real data set show the superiority of the ANT2FIS model in the prediction process with an average accuracy of 94% (AUC 99%) compared to other related works. These results are promising for early prediction of the natural disasters and prevention of its side effects.