边坡角设计的支持向量机建模与精度影响因素研究  

RESEARCH ON MODEL CONSTRUCTION OF SUPPORT VECTOR MACHINE AND PRECISION-INFLUEUCING FACTORS OF SLOPE ANGLE DESIGN

作　　者：刘开云[1] 乔春生[1] 田盛丰[2] 滕文彦[3] 

机构地区：[1]北京交通大学土木建筑工程学院,北京100044 [2]北京交通大学计算机与信息技术学院,北京100044 [3]石家庄铁路工程职业技术学院土木系,河北石家庄050041

出　　处：《岩石力学与工程学报》2005年第2期328-335,共8页Chinese Journal of Rock Mechanics and Engineering

基　　金：国家自然科学基金资助项目(50078002)

摘　　要：运用人工智能领域最新的基于结构风险最小化原理的机器学习算法——支持向量机(SVM)算法,采用线性Linear 和径向基函数 RBF 两种核函数以及ε 不敏感和 Quadratic 两种损失函数,并且考虑惩罚参数 C 的不同,编写了相应的程序对影响边坡角设计的诸多因素进行了机器学习,经过反复调整相关参数和计算对比,找到了拟合精度很高的支持向量机网络,并以此网络对测试样本作预测检验模型的可靠性;对影响支持向量机建模精度的各种影响因素作了计算和分析,在此基础上,初步确定了各参数对 SVM 模型精度影响大小的顺序,为 SVM 在类似工程上的应用提供了借鉴。Based on the structural risk minimization principle, a support vector machine (SVM) algorithm, the best machine learning algorithm in the artificial intelligence field today, is introduced. Two kinds of kernel functions( linear and radial basis function) and two kinds of loss functions (Ε-insensitive and quadratic) and different penalty parameter C are adopted to program a SVM routine in Matlab. Using the developed SVM model, many influencing factors of slope angle design are analyzed. With continued parameter modification and comparative calculations, a SVM network model with high accuracy of fitting was established. The reliability of this SVM network model is verified by sample testing, where many kinds of accuracy influencing factors of the SVM model are considered. The precision influencing sequence of these parameters is confirmed based on the calculation results, providing reference for similar engineering applications.

关 键 词：边坡工程 边坡角设计 支持向量机建模 机器学习与预测 参数分析 

分 类 号：TU413.62[建筑科学—岩土工程]