Scholarlink

<p>Frequent extreme heat events in urban areas have brought growing public attention to the urban heat island effect. However, traditional monitoring techniques are unable to handle high-dimensional data, which limits their application in real-world scenarios. This paper introduces SHapley Additive exPlanations to optimize Convolutional Neural Networks. The method uses min–max normalization to process big data and applies Gradient Descent to improve the model. It also analyzes the contribution of each heat island factor to the prediction results and generates interpretable evaluation outputs. A data set of urban heat island effect indices is used to validate the model. The results show that the model achieves a final harmonic mean of 0.93 and an accuracy of 88.7%. On a self-built data set of Xiamen’s heat island effect, the model reaches a Root Mean Square Error of 0.43&#xa0;°C, a mean prediction time of 12.3&#xa0;s, and a median of 11.8&#xa0;s. The prediction match rate of heat island intensity reaches 99.5%. These results confirm that the model is both reliable and efficient for monitoring and predicting the urban heat island effect. It can make accurate predictions across different urban areas and offers valuable insights for prevention strategies and supports the development of smart city.</p> Graphical Abstract <p></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

Urban heat island monitoring and intelligent prediction method based on big data analysis

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年9月19日

查看全文
  • Lin Wang

摘要

Frequent extreme heat events in urban areas have brought growing public attention to the urban heat island effect. However, traditional monitoring techniques are unable to handle high-dimensional data, which limits their application in real-world scenarios. This paper introduces SHapley Additive exPlanations to optimize Convolutional Neural Networks. The method uses min–max normalization to process big data and applies Gradient Descent to improve the model. It also analyzes the contribution of each heat island factor to the prediction results and generates interpretable evaluation outputs. A data set of urban heat island effect indices is used to validate the model. The results show that the model achieves a final harmonic mean of 0.93 and an accuracy of 88.7%. On a self-built data set of Xiamen’s heat island effect, the model reaches a Root Mean Square Error of 0.43 °C, a mean prediction time of 12.3 s, and a median of 11.8 s. The prediction match rate of heat island intensity reaches 99.5%. These results confirm that the model is both reliable and efficient for monitoring and predicting the urban heat island effect. It can make accurate predictions across different urban areas and offers valuable insights for prevention strategies and supports the development of smart city.

Graphical Abstract