Scholarlink

<p>Mountain flood in small watershed is widely distributed disaster, which have the characteristics of strong suddenness, great harm, and frequently. The traditional hydrodynamic and manual forecasting methods have high error rates for hourly forecasting. In order to improve the accuracy and real-time of water level forecasting in small watershed, we extract effective disaster-causing information, integrate multi-dimensional disaster-causing factors (such as hydrology, meteorology, geography, etc.), use a short-term prediction window and loop multi-step input method to improve the Machine Learning (ML) regression models’ accuracy, which can reduce the ML model’s process error. The non-ensemble and ensemble ML regression models is constructed for forecasting by loop multi-step, the non-ensemble models including Linear Regression (LR), Support Vector Machine Regression (SVMR) and <i>k</i>-Nearest Neighbors Regression (<i>k-</i>NNR), and the ensemble ML models include Random Forest Regression (RFR) and Gradient Boosting Regression (GBR). The loop multi-step ensemble ML regression models have the characteristics of high accurate and low time consumption than the general ML regression models for mountain flood forecasting in small watershed.</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

Mountain flood forecasting in small watershed based on loop multi-step machine learning regression model

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年4月11日

查看全文
  • Songsong Wang,
  • Bo Peng,
  • Ouguan Xu,
  • Yuntao Zhang,
  • Jun Wang

摘要

Mountain flood in small watershed is widely distributed disaster, which have the characteristics of strong suddenness, great harm, and frequently. The traditional hydrodynamic and manual forecasting methods have high error rates for hourly forecasting. In order to improve the accuracy and real-time of water level forecasting in small watershed, we extract effective disaster-causing information, integrate multi-dimensional disaster-causing factors (such as hydrology, meteorology, geography, etc.), use a short-term prediction window and loop multi-step input method to improve the Machine Learning (ML) regression models’ accuracy, which can reduce the ML model’s process error. The non-ensemble and ensemble ML regression models is constructed for forecasting by loop multi-step, the non-ensemble models including Linear Regression (LR), Support Vector Machine Regression (SVMR) and k-Nearest Neighbors Regression (k-NNR), and the ensemble ML models include Random Forest Regression (RFR) and Gradient Boosting Regression (GBR). The loop multi-step ensemble ML regression models have the characteristics of high accurate and low time consumption than the general ML regression models for mountain flood forecasting in small watershed.