基于时间序列的中国重要城市空气污染指数的建模与预测分析

Time Series Analysis of Air Pollution Indexes in Major Chinese Cities Modeling and Predictive Analytics

摘要: 中国城市化进程的持续推进，大型城市的空气污染已成为亟待解决的环境与公共健康问题。为提升空气质量预测的准确性，本文选取北京、上海、广州、天津及重庆五个国家级中心城市的2014~2024年的空气质量指数(AQI)逐日数据，构建季节性差分自回归滑动平均(SARIMA)模型与Holt-Winters指数平滑模型，并对二者的预测效能作对比检验。结果显示，两模型均能有效捕捉AQI序列的季节性与周期性波动，样本外预测的相对误差均低于10%；SARIMA模型在多数案例中表现更优的拟合精度。研究表明，城市产业能源结构及交通政策是影响空气质量变化的关键因素。

Abstract: Amidst ongoing urbanization in China, air pollution in major cities has become a pressing environmental and public health issue. To improve the accuracy of air quality forecasting, this study employs daily Air Quality Index (AQI) data from 2014 to 2024 for five national central cities—Beijing, Shanghai, Guangzhou, Tianjin, and Chongqing—to construct Seasonal Autoregressive Integrated Moving Average (SARIMA) and Holt-Winters exponential smoothing models, and compares their predictive performance. The results indicate that both models effectively capture seasonal and periodic fluctuations in AQI sequences, with out-of-sample prediction relative errors below 10%; however, the SARIMA model demonstrates better fitting accuracy in most cases. This study reveals that urban industrial-energy structures and transportation policies are key factors influencing air quality changes.