计算机工程与应用 ›› 2008, Vol. 44 ›› Issue (21): 145-147.DOI: 10.3778/j.issn.1002-8331.2008.21.040

• 机器学习 • 上一篇    下一篇

动态SVDD人机交互预警算法及其应用

彭敏晶1,2   

  1. 1.华南理工大学 工商管理学院,广州 510641
    2.五邑大学 系统科学与技术研究所,广东 江门 529020
  • 收稿日期:2008-04-30 修回日期:2008-05-29 出版日期:2008-07-21 发布日期:2008-07-21
  • 通讯作者: 彭敏晶

Dynamic SVDD based human-machine interactive early warning algorithm and its application

PENG Min-jing1,2   

  1. 1.School of Business Administration,South China University of Technology,Guangzhou 510641,China
    2.Institute of Systems Science and Technology,Wuyi University,Jiangmen,Guangdong 529020,China
  • Received:2008-04-30 Revised:2008-05-29 Online:2008-07-21 Published:2008-07-21
  • Contact: PENG Min-jing

摘要: 分析了当前主要的预警方法,指出了其不足,提出了动态SVDD人机交互预警算法。其中,人机交互技术的引入是为了让专家介入预警工作,以使系统能够准确地区分预警对象的状态。针对人机交互预警,分析了已有的SVDD技术,并指出随着训练样本数目的增加,该算法会因为过大的优化规模而无法操作。为此,提出动态SVDD算法,从而大大减小了优化规模,提高了人机交互预警系统的效率。

关键词: 人机交互, 预警, 动态SVDD, 支持向量, 核方法

Abstract: After reviewing the current early warning researches,it is found that current methods in these researches are unsuitable under some circumstances,and a human-machine interactive early warning algorithm based on dynamic support vector data description(SVDD) is proposed.In the algorithm,the technique of human-machine interactive can enable the early warning system accurately identify “good” and “bad” objects.And then the technique of SVDD is analyzed,and it is found that the human-machine early warning system couldn’t work because of the over large optimization scale.To solve the problem,the algorithm of dynamic SVDD is introduced to prompt the efficiency of the system by largely decreasing the optimization scale.

Key words: human-machine interaction, early warning, dynamic SVDD, support vector, kernel methods