基于样本权重的v-支持向量机  

v-Support vector machine with weight of samples

作　　者：李凯 翟璐璐 崔丽娟[2] LI Kai;ZHAI Lulu;CUI Lijuan(School of Cyber Security and Computer,Hebei University,Baoding 071002,China;Library,Hebei University,Baoding 071002,China)

机构地区：[1]河北大学网络空间安全与计算机学院,河北保定071002 [2]河北大学图书馆,河北保定071002

出　　处：《河北大学学报（自然科学版）》2018年第2期185-193,共9页Journal of Hebei University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61375075);河北省自然科学基金资助项目(F2018201060);河北大学自然科学研究计划项目(799207217074)

摘　　要：v-支持向量机克服了传统支持向量机选取参数的困难问题,然而,该方法并未考虑不同样本在训练中的作用,从而导致该方法对噪声或孤立点具有较强的敏感性,使得训练易出现过拟合现象.为了解决这些问题,针对不同样本的作用,通过引入样本的权重,获得了一种改进的v-支持向量机模型,使用Lagrange方法对该模型求解,获得了一种支持向量机分类器.实验中选取来自于UCI数据库的10个标准数据集,针对2种不同的确定样本权重的方法,验证了提出方法的有效性,并与C-SVM和v-SVM进行了性能比较.Theυ-support vector machine(υ-SVM)overcomes the difficult problem with selection of parameter for the traditional support vector machine.However,this method does not take into account the effect of different samples in the process of training,which leads to a strong sensitivity to noise or outlier points so that training is easy to be over-fit.In order to solve this problem,an improved v-SVM model is obtained by introducing the weight of the samples for the effect of different samples.A Lagrange method is used to solve the model,and a support vector machine classifier is obtained.To show the effectiveness of the presented method,the experiments are conducted using at two kinds of methods for determining weights of samples on 10 standard data sets from UCI repository,and compared with the performance of C-SVM andυ-SVM.

关 键 词：支持向量机 V-支持向量机 权重 模糊聚类 核策略 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]