I. Introduction
Nowadays, China's economy and industrialization have been rapidly growing, expanding urban populations and industrial scale. With the increased use of fossil fuels such as coal and oil, the levels of gaseous and solid pollutants in the air have also risen accordingly. Leading to air pollution getting worse and worse [1]. The air pollution in some areas has become so severe that it affects people's daily lives. Air pollution has attracted extensive attention from all countries. Scientific research has found that exposure to severely polluted air can cause many serious health problems in the human body, especially in vital parts such as the lungs and cardiovascular system [2], [3], [4], [5]. If the exposure time is too long, it may even cause cancer, damage the immune system, and even damage the reproductive system. Air pollution not only harms the ecological environment but also severely restricts economic development [6], [7]. The Air Quality Index (AQI) directly reflects air quality, with better air quality corresponding to a lower AQI [8], [9]. Accurate predictions of AQI can enhance the effectiveness of air pollution prevention and control. However, due to the dynamic and nonlinear nature of air distribution in the environment, analyzing and predicting AQI have proven particularly challenging [10], [11], [12]. As a result, researchers have devoted significant attention to exploring air data's non-stationarity and nonlinearity to improve AQI predictions' accuracy.