Scholarlink

<p>Climate change is intensifying the threats that natural disasters pose to global economic stability. To prepare for and respond to these disasters effectively, it's crucial to accurately predict the economic losses they will cause. This study introduces a new machine learning framework that incorporates future climate scenarios into disaster impact modeling. At its core is a K-Nearest Neighbors Regression (KNNR) scheme enhanced utilizing three bio-inspired algorithms: Tunicate Swarm Algorithm (TSA), Leader Harris Hawks Optimization (LHHO), and Seagull Optimization Algorithm (SOA). These algorithms enhance parameter selection and model efficiency. Among the hybrid models developed, the KNSO model showed the best performance, achieving a Root Mean Square Error (RMSE) of 1.90E+11 and an R<sup>2</sup> value of 0.995 during training. These results significantly outperform traditional models and demonstrate the effectiveness of combining evolutionary optimization with data-driven learning. Unlike traditional disaster models that rely solely on historical data, the proposed method accounts for changing climate conditions and future variability. The scheme offers a practical and scalable decision-support tool for policymakers aiming to reduce economic vulnerability, optimize resource allocation, and enhance resilience to future climate-related disasters.</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

Adaptive strategies for predicting economic impacts of natural disasters using hybrid optimization methods

  • 期刊论文

  • 发表日期：

    2025年8月11日

查看全文
  • Bing Zhao

摘要

Climate change is intensifying the threats that natural disasters pose to global economic stability. To prepare for and respond to these disasters effectively, it's crucial to accurately predict the economic losses they will cause. This study introduces a new machine learning framework that incorporates future climate scenarios into disaster impact modeling. At its core is a K-Nearest Neighbors Regression (KNNR) scheme enhanced utilizing three bio-inspired algorithms: Tunicate Swarm Algorithm (TSA), Leader Harris Hawks Optimization (LHHO), and Seagull Optimization Algorithm (SOA). These algorithms enhance parameter selection and model efficiency. Among the hybrid models developed, the KNSO model showed the best performance, achieving a Root Mean Square Error (RMSE) of 1.90E+11 and an R2 value of 0.995 during training. These results significantly outperform traditional models and demonstrate the effectiveness of combining evolutionary optimization with data-driven learning. Unlike traditional disaster models that rely solely on historical data, the proposed method accounts for changing climate conditions and future variability. The scheme offers a practical and scalable decision-support tool for policymakers aiming to reduce economic vulnerability, optimize resource allocation, and enhance resilience to future climate-related disasters.