基于SVDD的分类新方法

作者：杨振章，方景龙 日期：2009-12-30/span> 浏览：3681 查看PDF文档

基于SVDD的分类新方法

杨振章，方景龙
(杭州电子科技大学 计算机学院，浙江 杭州 310018）

摘要：为了实现对大规模数据集的分类，提出了一种基于支持向量数据描述(SVDD)的新算法。把两类的支持向量机（SVM）问题转化为软间隔的SVDD问题，然后用核集迭代的方法得到了近似最优解。实验结果表明，提出的算法与传统的SVM方法预测精度相当，但是训练时间和测试时间更短，产生的支持向量数也更少，适用于大规模数据的分类。
关键词：支持向量数据描述；核集；支持向量机
中图分类号：TP18文献标识码：A文章编号：1001-4551(2009)11-0048-04

A new classification method based on support vector data description
YANG Zhen-zhang, FANG Jing-long
(College of computer, Hangzhou Dianzi University, Hangzhou 310018, China)
Abstract: Aiming at realizing the classification of large-scale data sets，a new algorithm based on the support vector data description (SVDD) was proposed. The two-class support vector machine (SVM) were viewed as soft-margin support vector data description(SVDD) problem, and then approximately optimal solutions were obtained efficiently with the use of core sets. The results indicate that, it is as accurate as existing SVM implementations, but it is much faster and produces far fewer support vectors on large data sets.
Key words: support vector data description(SVDD); core sets; support vector machine(SVM)