北京邮电大学学报

  • EI核心期刊

北京邮电大学学报 ›› 2010, Vol. 33 ›› Issue (2): 20-23.doi: 10.13190/jbupt.201002.20.264

• 论文 • 上一篇    下一篇

双边界支持向量机的理论研究与分析

丁晓剑 赵银亮   

  1. (西安交通大学计算机科学与技术系, 西安 710049)
  • 收稿日期:2009-06-04 修回日期:2009-10-28 出版日期:2010-04-28 发布日期:2010-04-28
  • 通讯作者: 丁晓剑 E-mail:wjswl@163.com
  • 作者简介:丁晓剑(1982—), 男, 博士生, Email: wjswl@163.com; 赵银亮(1960—), 男, 教授, 博士生导师.
  • 基金资助:

    国家高技术研究发展计划项目(2008AA01Z136

Theory and Analysis of DoubleMargin SVM

DING Xiao-jian, ZHAO Yinl-iang   

  1. (Department of Computer Science and Technology, Xi'an Jiaotong University, Xi'an 710049, China)
  • Received:2009-06-04 Revised:2009-10-28 Online:2010-04-28 Published:2010-04-28
  • Contact: DING Xiao-jian E-mail:wjswl@163.com

摘要:

根据统计学习理论,间隔大小是反映泛化能力的一个很重要的方面. 受一类支持向量机(SVM)的启发,提出的双边界SVM能分别用2个边界对2类问题分类. 它能在保证分类正确的同时保证分类间隔的最大化,理论上分别从推广性能和不平衡类分布2方面证明了其优越性. 标准数据集上的实验表明,双边界SVM得到的分类间隔要大于SVM, 泛化性有了显著提高;另外,不平衡数据集上分析得到它对少数类识别率有明显提升. 真实入侵数据测试结果表明,双边界SVM算法比边界样本选择算法的检测率高出2%以上.

关键词: 分类间隔, 泛化性能, 双边界支持向量机

Abstract:

Based on the statistical learning theory (SLT), the margin scale reflects the generalization capability to a great extent. Inspired by oneclass support vector machine (SVM), doublemargin SVM is put forward to classify two classes by two margins separately. Instances can be classified correctly as well as margin maximization, and its superiority is theoretical proved by both generalization performance and imbalanced class distribution. Experiment on benchmark data sets shows that classification margin obtained by doublemargin SVM is larger than SVM, improving the generalization apparently, and analysis on imbalanced data shows that it has a higher recognition ratio. Finally real intrusion detection data shows that the detection precision is increased by 2% against boundary samples selection method.

Key words: classification margin, generalization capability, doublemargin support vector machine