摘要
从时频分析角度出发,探讨利用小波分析与LSSVM模型作滑坡变形预测。其步骤为利用小波变换把变形时间序列分解成具有不同频率特征的分量,再对重构后的近似序列和细节序列分别利用LSSVM进行预测并将结果融合。实例结果表明,基于小波分析与LSSVM的滑坡变形预测方法预测精度高于GM(1,1)、AR和单一的LSSVM方法。
A novel model based on wavelet analysis and Least Square Support Vector Machine （LSSVM） for landslide deformation prediction is presented. Firstly, in the view of time-frequency analysis, through the wavelet transform, deformation time series is decomposed into components of different frequency and then the reconstructed approximate series and detailed series were predicted respectively by using LSSVM and the results were composed finally. The experimental results indicates that this prediction model has advantage over GM （ 1,1 ） , AR and simple LSSVM as it has higher prediction accuracy and is applicable to predicting landslide deformation.
出处
《大地测量与地球动力学》
CSCD
北大核心
2009年第4期127-130,共4页
Journal of Geodesy and Geodynamics
基金
国家自然科学基金(40674008)
关键词
小波分析
最小二乘支持向量机
滑坡变形预测
时间序列
细节序列
wavelet analysis
least square support vector machine （LSSVM）
landslide deformation prediction
time series
detailed series
