Scholarlink

<p>Effective flood probability prediction is the backbone of disaster risk management and, consequently, hazard mitigation against flood events in both communities and ecosystems. This article presents the latest developments in predictive modeling using two robust approaches: CatBoost (Categorical Boosting) regression and Multilayer Perceptron regression. To enhance predictive performance, innovative hybrid models were developed by integrating the African Vultures Optimization Algorithm with these machine learning methods. Model performance was evaluated using five key metrics, including Root Mean Square Error, R-squared (R<sup>2</sup>), and Root Standard Ratio (RSR). Among all tested models, the CatBoost African Vultures (CAAV) hybrid model attained the highest predictive accuracy, with an R<sup>2</sup> value of 0.988 and an RMSE of 0.006. These findings highlight the significant potential of optimized hybrid schemes in accurate flood prediction. Beyond their academic value, the results offer practical implications for real-world applications in flood risk management, early warning systems, and urban planning strategies aimed at minimizing the socioeconomic and environmental impacts of flooding. This study opens new avenues for combining machine learning and biologically inspired optimization to tackle critical environmental challenges.</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

Optimizing the performance of novel intelligent models by integrating metaheuristic algorithms for flood probability estimation

  • 期刊论文

  • 发表日期：

    2025年7月30日

查看全文
  • Pengju Zhang,
  • Yujuan Shi,
  • Xiaowen Ren,
  • Yongjie Shi,
  • Jianxin Zhang

摘要

Effective flood probability prediction is the backbone of disaster risk management and, consequently, hazard mitigation against flood events in both communities and ecosystems. This article presents the latest developments in predictive modeling using two robust approaches: CatBoost (Categorical Boosting) regression and Multilayer Perceptron regression. To enhance predictive performance, innovative hybrid models were developed by integrating the African Vultures Optimization Algorithm with these machine learning methods. Model performance was evaluated using five key metrics, including Root Mean Square Error, R-squared (R2), and Root Standard Ratio (RSR). Among all tested models, the CatBoost African Vultures (CAAV) hybrid model attained the highest predictive accuracy, with an R2 value of 0.988 and an RMSE of 0.006. These findings highlight the significant potential of optimized hybrid schemes in accurate flood prediction. Beyond their academic value, the results offer practical implications for real-world applications in flood risk management, early warning systems, and urban planning strategies aimed at minimizing the socioeconomic and environmental impacts of flooding. This study opens new avenues for combining machine learning and biologically inspired optimization to tackle critical environmental challenges.