Novel analysis–forecast system based on multi-objective optimization for air quality index

Abstract

The air quality index (AQI) is an important indicator of air quality. Owing to the randomness and non-stationarity inherent in AQI, it is still a challenging task to establish a reasonable analysis–forecast system for AQI. Previous studies primarily focused on enhancing either forecasting accuracy or stability and failed to improve both aspects simultaneously, leading to unsatisfactory results. In this study, a novel analysis–forecast system is proposed that consists of complexity analysis, data preprocessing, and optimize–forecast modules and addresses the problems of air quality monitoring. The proposed system performs a complexity analysis of the original series based on sample entropy and data preprocessing using a novel feature selection model that integrates a decomposition technique and an optimization algorithm for removing noise and selecting the optimal input structure, and then forecasts hourly AQI series by utilizing a modified least squares support vector machine optimized by a multi-objective multi-verse optimization algorithm. Experiments based on datasets from eight major cities in China demonstrated that the proposed system can simultaneously obtain high accuracy and strong stability and is thus efficient and reliable for air quality monitoring.