基于网格搜索的支持向量机砂土液化预测模型
|
摘要
在使用支持向量机对砂土液化进行预测时,不可避免地会遇到惩罚因子和核函数参数如何选取的问题,目前还没有确定这两个参数的选择模式,解决这一问题比较常用的办法有经验公式法、遗传算法和网格搜索法。对此本文基于网格搜索方法,使用LIBSVM软件包,寻找砂土液化训练样本的结构风险最小值所对应的支持向量机最优参数;使用最优参数,建立了支持向量机砂土液化预测模型。研究结果表明:预测样本的正确率可达87.5%,而且预测结果稳定;同时通过比较BP神经网络的砂土液化预测情况,可知支持向量机有更好的泛化能力,收敛速度也更快。
When the support vector machine is used for predicting sand liquefaction,the problem how to choose penalty factor and parameter of kernel functionis inevitably encountered,and the usual method for solving this problem is the empirical formula method,the genetic algorithm method and the grid-search method.This paper uses software called LIBSVM and the grid-search method to search parameters of the minimum structure risk of SVM through training the sample of sand liquefaction.Then,a predicting model of SVM with parameters,which have been obtained above,is set up to predict the sample of prediction,and the results show that the correct rate is 87.5%.Through comparing with the prediction of sand liquefaction by BP Network,the support vector machine shows better generalization ability and faster convergence speed.
When the support vector machine is used for predicting sand liquefaction,the problem how to choose penalty factor and parameter of kernel functionis inevitably encountered,and the usual method for solving this problem is the empirical formula method,the genetic algorithm method and the grid-search method.This paper uses software called LIBSVM and the grid-search method to search parameters of the minimum structure risk of SVM through training the sample of sand liquefaction.Then,a predicting model of SVM with parameters,which have been obtained above,is set up to predict the sample of prediction,and the results show that the correct rate is 87.5%.Through comparing with the prediction of sand liquefaction by BP Network,the support vector machine shows better generalization ability and faster convergence speed.
引文
[1]谢定义.土动力学[M].西安:西安交通大学出版社,1988.
[2]刘金韬,金晓媚.饱和砂土液化判别方法中问题浅析[J].地质工程学报,2000,8(3):379-384.
[3]Seed H B,Idriss I M.Simpliifed procdeure for evaluating soil liquefaction potenital[J].J of Geot Eng Div,ASCE,1971,97(SM9):1249-l273.
[4]杨秀竹,王星华,雷金山.地震时饱和砂土液化机理及统计判别法[J].长沙铁道学院学报,2001,19(1):35-37.
[5]刘章军,叶燎原,彭刚.砂土地震液化的模糊概率评判方法[J].岩土力学,2008,29(4):876-880.
[6]潘健,刘利艳,林慧常.基于BP神经网络的砂土液化影响因素的综合评估[J].华南理工大学学报:自然科学版,2006,34(11):76-79.
[7]赵艳林,杨绿峰,吴敏哲.砂土液化的灰色综合评判[J].自然灾害学报,2000,9(1):72-78.
[8]夏建中,罗战友,龚晓南,等.基于支持向量机的砂土液化预测模型[J].岩石力学与工程学报,2005,24(22):4139-4144.
[9]陈荣淋,林建华,黄群贤.支持向量机在砂土液化预测中的应用研究[J].中国地质灾害与防治学报,2005,16(2):15-23.
[10]邓乃扬,田英杰.数据挖掘中的新方法:支持向量机[M].北京:科学出版社,2004.
[11]Vapnik V N.The nature of statistical learning[M].New York,Springer,1995.
[12]Vapnik V N.An overview of statistical learning theory[J].IEEE Trans on Neural Network,1999,10(5):988-999.
[13]赵洪波.岩土力学与工程中的支持向量机分析[M].北京:煤炭工业出版社,2008.
[14]唐山地震砂土液化联合研究小组.唐山地震砂土液化现场勘察资料研究报告[R].北京:北京市勘察处,1983,2:26-41.
[15]李明,陈国兴.基于BP神经网络的饱和砂土液化判别方法[J].自然灾害学报,2005,14(2):108-114.
[2]刘金韬,金晓媚.饱和砂土液化判别方法中问题浅析[J].地质工程学报,2000,8(3):379-384.
[3]Seed H B,Idriss I M.Simpliifed procdeure for evaluating soil liquefaction potenital[J].J of Geot Eng Div,ASCE,1971,97(SM9):1249-l273.
[4]杨秀竹,王星华,雷金山.地震时饱和砂土液化机理及统计判别法[J].长沙铁道学院学报,2001,19(1):35-37.
[5]刘章军,叶燎原,彭刚.砂土地震液化的模糊概率评判方法[J].岩土力学,2008,29(4):876-880.
[6]潘健,刘利艳,林慧常.基于BP神经网络的砂土液化影响因素的综合评估[J].华南理工大学学报:自然科学版,2006,34(11):76-79.
[7]赵艳林,杨绿峰,吴敏哲.砂土液化的灰色综合评判[J].自然灾害学报,2000,9(1):72-78.
[8]夏建中,罗战友,龚晓南,等.基于支持向量机的砂土液化预测模型[J].岩石力学与工程学报,2005,24(22):4139-4144.
[9]陈荣淋,林建华,黄群贤.支持向量机在砂土液化预测中的应用研究[J].中国地质灾害与防治学报,2005,16(2):15-23.
[10]邓乃扬,田英杰.数据挖掘中的新方法:支持向量机[M].北京:科学出版社,2004.
[11]Vapnik V N.The nature of statistical learning[M].New York,Springer,1995.
[12]Vapnik V N.An overview of statistical learning theory[J].IEEE Trans on Neural Network,1999,10(5):988-999.
[13]赵洪波.岩土力学与工程中的支持向量机分析[M].北京:煤炭工业出版社,2008.
[14]唐山地震砂土液化联合研究小组.唐山地震砂土液化现场勘察资料研究报告[R].北京:北京市勘察处,1983,2:26-41.
[15]李明,陈国兴.基于BP神经网络的饱和砂土液化判别方法[J].自然灾害学报,2005,14(2):108-114.
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心