Scholarlink

<p>Driven by the dual policies of accelerated global environmental governance and low-carbon transition, PM<sub>2.5</sub>, owing to its high diffusivity and concentration, poses a serious threat to public health and regional air quality. Accordingly, developing a low-complexity, high-accuracy prediction system has become a key approach to strengthening air quality management and ecological security. Yet addressing PM<sub>2.5</sub> data, intertwined with multiple factors and characterized by seasonality and nonlinear dynamics, remains a pressing challenge in current prediction research. This study proposes a low-complexity PM<sub>2.5</sub> prediction system based on multi-objective ensemble optimization. The system decomposes quarterly PM<sub>2.5</sub> time series into multiple intrinsic mode functions via a decomposition strategy, suppresses noise according to the variance contribution rate, and inputs retained key features into the baseline model optimized by the Crested Porcupine Optimization algorithm. Furthermore, by applying the multi-objective version of this algorithm, with dual objectives of minimizing mean absolute percentage error and prediction variance, the baseline model’s weight distribution is globally searched and optimized, thereby yielding a system with superior performance. On this basis, the clustering non-parametric Bootstrap method clusters and resamples point estimates, generating multiple datasets and providing interval estimates across confidence levels, thus quantifying model uncertainty. An empirical analysis using fourth-quarter data from the Beijing–Tianjin–Hebei region in China shows that the proposed model enhances prediction accuracy and interpretability, strengthens adaptability and generalization, overcomes seasonal barriers in air pollution forecasting, and advances environmental governance and low-carbon policymaking toward scientific, streamlined, and intelligent orientations.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A low-complexity ensemble prediction model for PM2.5 from the perspective of global environmental governance: a seasonal empirical study in the Beijing–Tianjin–Hebei Region of China

  • 期刊论文

  • 发表日期：

    2026年3月25日

查看全文
  • Xiayan Liu,
  • Bo Zeng,
  • Xinjie Shi,
  • Jianzhou Wang

摘要

Driven by the dual policies of accelerated global environmental governance and low-carbon transition, PM2.5, owing to its high diffusivity and concentration, poses a serious threat to public health and regional air quality. Accordingly, developing a low-complexity, high-accuracy prediction system has become a key approach to strengthening air quality management and ecological security. Yet addressing PM2.5 data, intertwined with multiple factors and characterized by seasonality and nonlinear dynamics, remains a pressing challenge in current prediction research. This study proposes a low-complexity PM2.5 prediction system based on multi-objective ensemble optimization. The system decomposes quarterly PM2.5 time series into multiple intrinsic mode functions via a decomposition strategy, suppresses noise according to the variance contribution rate, and inputs retained key features into the baseline model optimized by the Crested Porcupine Optimization algorithm. Furthermore, by applying the multi-objective version of this algorithm, with dual objectives of minimizing mean absolute percentage error and prediction variance, the baseline model’s weight distribution is globally searched and optimized, thereby yielding a system with superior performance. On this basis, the clustering non-parametric Bootstrap method clusters and resamples point estimates, generating multiple datasets and providing interval estimates across confidence levels, thus quantifying model uncertainty. An empirical analysis using fourth-quarter data from the Beijing–Tianjin–Hebei region in China shows that the proposed model enhances prediction accuracy and interpretability, strengthens adaptability and generalization, overcomes seasonal barriers in air pollution forecasting, and advances environmental governance and low-carbon policymaking toward scientific, streamlined, and intelligent orientations.