基于VMD-SegSigmoid-XGBoost-ClusterLSTM算法的山体滑坡表面位移预测  

Prediction of Surface Displacement of Landslides Based on VMD-SegSigmoid-XGBoost-ClusterLSTM Algorithm

作　　者：李瑞晨 侯木舟[1] 孔梦麟 谢昊含 Li Ruichen;Hou Muzhou;Kong Menglin;Xie Haohan(School of Mathematics and Statistics,Central South University,Changsha 410083,China;School of Earth Science and Information Physics,Central South University,Changsha 410013,China)

机构地区：[1]中南大学数学与统计学院,长沙410083 [2]中南大学地球科学与信息物理学院,长沙410013

出　　处：《科技通报》2024年第9期111-115,共5页Bulletin of Science and Technology

摘　　要：山体滑坡表面位移的预测可以帮助预估新的潜在滑裂面,避免造成更加严重的危害。本文针对芷江县禾梨坳乡大沙界村牛塘坳组滑坡单方向表面位移数据进行建模研究,提出一种基于变分模态分解的时间序列预测框架VMD-SegSigmoid-XGBoost-ClusterLSTM,可较准确地预测滑坡表面位移。该模型在数据集上表现较好,除去较难拟合的残差项子序列,趋势项子序列和周期项子序列的均方根误差和平均绝对百分比误差均小于0.1,其中XGBoost周期项预测模块的均方根误差低至0.006。Predicting the surface displacement of mountain landslides can help predict new potential sliding surfaces and avoid causing more serious harm.This article conducts modeling research on the one-way surface displacement data of landslides in Niutangao Formation,Dashajie Village,Heliao Township,Zhijiang County,and proposes a time series prediction framework based on variational mode decomposition,VMD-SegSigmoid-XGBoost-ClusterLSTM,which can accurately predict the surface displacement of landslides.The model performs well in the data set,the rootmean-square deviation and average absolute percentage error of trend subsequence and periodic subsequence are both less than 0.1,except for the residual subsequence that is difficult to fit,the root-mean-square deviation of XGBoost periodic prediction module is as low as 0.006.

关 键 词：滑坡预警 表面位移 模态分解 时间序列 机器学习 

分 类 号：X4[环境科学与工程—灾害防治]