Scholarlink

<p>Forecasting air pollution concentrations, with a particular focus on particulate matter (PM), is a critical component of environmental monitoring systems, enabling timely warnings of smog episodes and the implementation of preventive measures to protect public health. This study introduces a hybrid forecasting approach that combines a transformer-based deep learning model with symbolic regression to predict weather-dependent PM concentrations. The transformer architecture is used to capture complex temporal dependencies in meteorological variables influencing PM variability, while symbolic regression provides interpretable mathematical expressions linking meteorological conditions to PM levels. The proposed method was evaluated using data from multiple air quality monitoring stations, achieving high predictive accuracy while maintaining model transparency, and was compared with several state-of-the-art baseline models. The results demonstrate the efficacy of the proposed approach, achieving an average MAE of approximately 0.10, compared to over 0.20 for the baseline models, an MSE of around 0.02 versus 0.08, and an RMSE of 0.15 compared to 0.25–0.30 for competing methods. These results confirm that the proposed hybrid model delivers substantially more accurate forecasts while preserving interpretability.</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 hybrid transformer and symbolic regression model for weather-dependent particulate matter forecasting in air quality management

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月19日

查看全文
  • Jarosław Bernacki

摘要

Forecasting air pollution concentrations, with a particular focus on particulate matter (PM), is a critical component of environmental monitoring systems, enabling timely warnings of smog episodes and the implementation of preventive measures to protect public health. This study introduces a hybrid forecasting approach that combines a transformer-based deep learning model with symbolic regression to predict weather-dependent PM concentrations. The transformer architecture is used to capture complex temporal dependencies in meteorological variables influencing PM variability, while symbolic regression provides interpretable mathematical expressions linking meteorological conditions to PM levels. The proposed method was evaluated using data from multiple air quality monitoring stations, achieving high predictive accuracy while maintaining model transparency, and was compared with several state-of-the-art baseline models. The results demonstrate the efficacy of the proposed approach, achieving an average MAE of approximately 0.10, compared to over 0.20 for the baseline models, an MSE of around 0.02 versus 0.08, and an RMSE of 0.15 compared to 0.25–0.30 for competing methods. These results confirm that the proposed hybrid model delivers substantially more accurate forecasts while preserving interpretability.