Author(s): Meiman Zhang; Youming Zhang; Jin Xu; Tianyu Lei; Mengtian Wu; Hai Zhu; Lingling Wang
Linked Author(s): Lingling Wang
Keywords: Dam break RFE algorithm SHAP method feature selection machine learning model
Abstract: Dam break prediction is important to amplify the precision and expeditiousness of the dam break process and disaster relief downstream. However, there are redundant dam break parameters, influencing the accuracy of models. This study obtains the sequence of the importance of features affecting peak discharge based on SHAP method. RFE algorithm is used to obtain respective optimal features in six ML models to reduce irrelevant and redundant parameters. This study proposes six ML models to predict peak discharge for breach with respective optimal features. And the evaluation result shows that XGBoost has the best performance. This study can provide technical support for the prediction of dam breach, provide guidance for emergency rescue work, and help to reduce the disaster loss of dam breach.
Year: 2025