两种不确定支持向量机分类性能的对比研究  

Comparative Study on Classification Performances of Two Indeterminate Support Vector Machines

作　　者：刘成忠[1] 

机构地区：[1]甘肃农业大学信息科学技术学院,甘肃兰州730070

出　　处：《计算机技术与发展》2011年第11期156-159,共4页Computer Technology and Development

基　　金：甘肃省自然科学基金(096RJZA004);甘肃省教育科研基金(0902-04);甘肃省科技支撑计划(1011NKCA058)

摘　　要：为了克服支持向量机方法对于噪声或孤立野值点敏感的问题,通过引入模糊理论与粗糙集方法,可以分别得到两种不确定支持向量机模型。文中通过分析和比较模糊支持向量机和粗糙支持向量机分类模型构造方法,解释了这两种不确定支持向量机模型克服噪声影响的原理。同时通过一个合成数据集和一组标准数据集对这两种不确定支持向量机的泛化性能进行了对比验证。实验结果表明,相比传统支持向量机,两种不确定支持向量机都能不同程度地提高分类精度,并且模糊支持向量机算法整体表现出了更好的泛化性能。In order to overcome the problem that support vector machine is ,sensitive to the noise and isolated outliers, introduce fuzzy theory and rough set theory into support vector machine to get two kinds of indeterminate support vector machines. Through analysis and comparison of the construction method of fuzzy support vector raachine and that of rough support vector machine, the principles of the two indeterminate methods reducing the outliers are explained. At the same time, generalization performances of the two indeterminate support vector machines are comparatively verified through a synthetic data set and a set of standard data. Experiment remits show that the two indeterminate methods have better performances of reducing oufliers than traditional support vector machine, that they can significantly improve the classification accuracy, and that fuzzy support vector machine has a better generalization performance oh the whole.

关 键 词：支持向量机 模糊理论 粗糙集 

分 类 号：TP31[自动化与计算机技术—计算机软件与理论]