计算机工程与应用 ›› 2008, Vol. 44 ›› Issue (3): 183-185.

• 数据库与信息处理 • 上一篇    下一篇

一种提高支持向量机性能的特征选择新方法

孙 刚1,2,王志平1,王明新3   

  1. 1.大连海事大学 数学系,辽宁 大连 116026
    2.通化师范学院 数学系,吉林 通化 134002
    3.北华大学 理学院,吉林 132000
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2008-01-21 发布日期:2008-01-21
  • 通讯作者: 孙 刚

New feature selection method of improve performance of SVM

SUN Gang1,2,WANG Zhi-ping1,WANG Ming-xin3   

  1. 1.Department of Mathematics,Dalian Maritime University,Dalian,Liaoning 116026,China
    2.Department of Mathematics,Tonghua Normal University,Tonghua,Jilin 134002,China
    3.College of Sciences,Beihua University,Jilin 132000,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2008-01-21 Published:2008-01-21
  • Contact: SUN Gang

摘要: 利用支持向量机进行模式分类时,特征选择是数据预处理的一项重要内容。有效的特征选择在很大程度上影响着分类器的性能。根据样本各特征分量的均值与方差对分类的影响,提出根据分类权值进行特征选择,以提高支持向量机性能的简便方法,制定了两个具体实施方案。在三个常用数据集上进行了仿真实验,结果验证了方法的有效性。

关键词: 支持向量机, 特征选择, 分类权

Abstract: Feature selection is an important content of data preprocessing in pattern recognition based on SVM.The valid feature selection affects the performance of classification machine to a large extent.According to the effect of the mean and the square difference of each feature of samples,a simple and convenient feature selection method based on the values of classification power is presented,and two concrete schemes are introduced to improve the performance of support vector machine.Three experiments on common datasets confirm the usefulness of the method.

Key words: Support Vector Machine(SVM), feature selection, classification power