Evolutionary Techniques for Optimizing Air Quality Model
Rashmi Bhardwaj, Dimple Pruthi
Abstract
In this study, the adaptive neuro-fuzzy inference system (ANFIS) is used to perform predictive analysis of air pollutant-fine particulate matter. Fine particulate matter are of size less than or equal to 2.5 microns (PM2.5) which is the prime pollutant because of its severe effects on pulmonary and nervous systems. The shortcomings of ANFIS to get trapped in local optima or computational complexity lead to the introduction of algorithms inspired by nature. The evolutionary algorithms form on flocking and genetics are employed in artificial intelligence. Many attempts have been made and are still going on to improve the accuracy in forecasting air quality. The study contributes to the advancing research by examining pertinence of evolutionary algorithms-particle swarm optimization (PSO) and genetic algorithm (GA) with ANFIS in predicting fine particulate matter. The non-stationary PM2.5 time series is decomposed using the discrete wavelet transform and sub-band predictive analysis is carried out. The wavelet ANFIS-PSO method is also compared with wavelet ANFIS (WANFIS), WANFIS-GA and classic ANFIS. The analysis revealed that WANFIS-PSO outperforms the other models.