Scholarlink

<p>Understanding sediment transport is fundamental for sustainable river management, hydraulic structure design, and ecological preservation. Despite several empirical, numerical, and experimental investigations, accurate prediction of bed load transport remains a persistent challenge due to the complex interactions of flow hydraulics and sediment properties. To address this gap, the present study evaluates the applicability of machine learning algorithms for predicting bed load transport discharge using a comprehensive dataset of hydraulic and sediment parameters. Four models Random Forest, Extra Trees, Extreme Gradient Boosting, and CatBoost were developed and compared using statistical performance indicators such as RMSE, NSE, PCC and R<sup>2</sup>. Among the tested models, XGBoost has achieved the highest prediction accuracy, demonstrating its robustness in capturing nonlinear relationships governing the sediment transport. The results highlight the potential of machine learning approaches as reliable alternatives to conventional empirical and numerical formulations, thereby contributing to improved sediment management and hydraulic engineering practices.</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

SHAP-based interpretation of machine learning models in bed load transport prediction

  • 期刊论文

  • 发表日期：

    2025年12月26日

查看全文
  • Mun Mun Basumatary,
  • Soumen Maji,
  • Bimlesh Kumar

摘要

Understanding sediment transport is fundamental for sustainable river management, hydraulic structure design, and ecological preservation. Despite several empirical, numerical, and experimental investigations, accurate prediction of bed load transport remains a persistent challenge due to the complex interactions of flow hydraulics and sediment properties. To address this gap, the present study evaluates the applicability of machine learning algorithms for predicting bed load transport discharge using a comprehensive dataset of hydraulic and sediment parameters. Four models Random Forest, Extra Trees, Extreme Gradient Boosting, and CatBoost were developed and compared using statistical performance indicators such as RMSE, NSE, PCC and R2. Among the tested models, XGBoost has achieved the highest prediction accuracy, demonstrating its robustness in capturing nonlinear relationships governing the sediment transport. The results highlight the potential of machine learning approaches as reliable alternatives to conventional empirical and numerical formulations, thereby contributing to improved sediment management and hydraulic engineering practices.