摘要
针对SVM在大类别模式分类中存在的问题,提出了一种基于模糊核聚类的SVM多类分类方法,并给出了一种高效的半模糊核聚类算法。该方法基于模糊核聚类方法生成模糊类,并采用树结构将多个SVM组合起来实现多类分类。模糊核聚类方法不但能够实现更为准确的聚类,而且能够挖掘模糊类的外围、不同模糊类之间的交叠情况等信息,利用这些信息能有效提高分类器的性能。实验表明,所提方法比传统方法具有更高的速度和精度。
Aimed at the problems of support vector machines（SVM） for multi-class pattern recognition with large number of catalogs, a new method of SVM multi-class classification based on fuzzy kernel clustering is proposed. In addition, an efficiency semi-fuzzy kernel clustering algorithm is presented. The new method defines confusion classes based on fuzzy kernel clustering and builds binary trees of support vector machines for the multi-class classification. The fuzzy kernd clustering can not only obtain a better performance than classical clustering, but also provide information about the boundary of a confusion class and the overlap between classes. The performance of support vector machines can be improved efficiently by using the information, Experimental results indicate that the new method yields higher precision and speed than classical SVM multi-class classification methods.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2006年第5期770-774,共5页
Systems Engineering and Electronics
基金
国家自然科学基金重点项目资助课题(70431001)
关键词
支持向量机
多类分类
模糊核聚类
树型分类器
support vector machine
multi-class classification
fuzzy kernel clustering
tree classifier
