利用改进水循环优化SVR的边坡变形预测  

Slope Deformation Prediction Based on IWCA-SVR

作　　者：董延东 DONG Yandong(Guoneng Transportation Technology Research Institute Co.,Ltd.,Beijing 100080,China)

机构地区：[1]国能运输技术研究院有限责任公司,北京100080

出　　处：《地理空间信息》2024年第10期46-49,共4页Geospatial Information

基　　金：国家能源集团科技创新项目(GJNY-20-231);国能朔黄铁路公司科技创新项目([2021]367)。

摘　　要：边坡变形与地质活动,气候变化以及人工切坡等众多因素相关,是一种典型的非线性、非平稳随机过程,传统边坡变形预测方法存在预测精度低和泛化能力弱等问题。提出一种基于改进水循环算法(improved water cycle algorithm,IWCA)优化支持向量回归(suppor vector regression,SVR)的边坡变形预测方法(IWCA-SVR)。SVR利用核函数将低维空间中的非线性边坡位移量数据映射到高维空间进行建模分析,不仅能够获得较高的预测精度,同时具有较强的泛化能力。由于SVR核参数和惩罚因子对预测性能影响较大,提出IWCA算法对其进行全局寻优,提升预测性能。最后将所提IWCA-SVR方法与粒子群优化的SVR(PSO-SVR)和极限学习机(ELM)等方法进行对比,结果表明IWCA-SVR在平均相对误差和均方根误差两项指标方面分别提升超过55.8%和54.9%,并且具有更强的泛化能力。Slope deformation is related to many factors such as geological activity,climate change and artificial slope cutting.Slope deformation is a typical nonlinear and non-stationary stochastic process.The traditional slope deformation prediction model has the problems of low prediction accuracy and weak generalization ability.We proposed a slope deformation prediction method based on improved water cycle algorithm optimized support vector regression(IWCA-SVR).SVR uses kernel function to map nonlinear slope displacement data in low-dimensional space to high-dimensional space for modeling and analysis,which can not only obtain high prediction accuracy,but also have strong generalization ability.SVR kernel parameters and penalty factors have a great impact on prediction performance.So we used IWCA algorithm to optimize it globally to improve prediction performance.Finally,we compared IWCA-SVR method with particle swarm optimization SVR and limit learning machine.The results show that the average relative error and root mean square error of IWCA-SVR have increased by more than 55.8%and 54.9%respectively,and it has stronger generalization ability.

关 键 词：铁路边坡 位移预测 水循环算法 支持向量回归 泛化能力 

分 类 号：P258[天文地球—测绘科学与技术]