随机森林优化的静动态耦合模型在滑坡位移预测中的应用
Application of static dynamic coupling model optimized by random forest in landslide displacement prediction
蒋宏伟 1刘健鹏 1王新杰 1陈春红 1刘惠1
作者信息
- 1. 常州大学城市建设学院,江苏常州 213164
- 折叠
摘要
以重庆市奉节县生基包滑坡为例,首先采用静态的支持向量回归(SVR)机器学习算法和动态的长短期记忆神经网络(LSTM)机器学习算法对滑坡位移进行预测;其次引入随机森林(RF)算法,在输入因素筛选的基础上,对SVR模型和LSTM模型的预测结果进行更优解分类预测;最后通过RF模型输出概率值,对静动态耦合模型(SVR-LSTM)进行权重赋值,得到RF优化的SVR-LSTM滑坡位移预测模型.结果表明LSTM模型预测整体优于SVR模型,RF优化的SVR-LSTM滑坡位移预测模型集成了静态SVR与动态LSTM预测模型的优势,其预测性能与单一的SVR模型和LSTM模型相比更优.研究提供了一种滑坡位移预测模型集成的思路,为三峡库区的地质灾害预测预报提供借鉴和参考.
Abstract
This paper took the Shengjibao landslide in Fengjie county,Chongqing as an example.A static machine learning algorithm called the support vector regression(SVR)and a dynamic machine learning algorithm called the long short-term memory neural network(LSTM)were proposed to pre-dict the landslide displacement.Then,the random forest(RF)algorithm was introduced to classify and predict the optimal solution between the SVR model and the LSTM model.Finally,the RF-opti-mized SVR-LSTM landslide displacement prediction model was obtained by assigning weights to the static-dynamic coupling model(SVR-LSTM)based on the probability values of the output from the RF model.The results show that LSTM model has better performance than the SVR model.RF-optimized SVR-LSTM landslide displacement prediction model integrates the advantages of static and dynamic prediction models,and its prediction performance is better than that of the SVR model and the LSTM model,respectively.This study provides an idea of integrating landslide displacement prediction model,which can provide reference for geological disaster prediction in the Three Gorges Reservoir area.
关键词
滑坡位移预测/随机森林/长短期记忆神经网络/支持向量回归/算法集成Key words
landslide displacement prediction/random forest/long short-term memory neural net-work/support vector regression/algorithm integrated引用本文复制引用
基金项目
常州大学人才引进资助项目(ZMF22020036)
出版年
2024
国家科技期刊平台
NSTL
万方数据