一种设计层次支持向量机多类分类器的新方法-计算机应用研究.PDFVIP

·34 · 计算机应用研究 2006 年 一种设计层次支持向量机多类分类器的新方法 * 1 1 2 赵 晖 , 荣莉莉 , 李 晓 ( 1. 大连理工大学 系统工程研究所, 辽宁 大连 116024; 2. 中国科学院 新疆理化技术研究所, 新疆 乌鲁木齐 830011) 摘 要: 层次结构的设计是层次支持向量机多类分类方法应用中的关键问题, 类间可分性是设计层次结构的重 要依据, 提出了一种基于线性支持向量机度量类间相似程度的方法 , 并给出了一种基于类间可分性设计层次支 持向量机多类分类器的新方法。实验表明, 新方法有效地提高了层次支持向量机多类分类器的分类精度和速 度。 关键词: 支持向量机; 多类分类; 层次结构 ; 类间可分性 中图法分类号: TP391 文献标识码 : A 文章编号 : 1001- 3695( 2006) 06- 0034- 04 New Method of Design Hierarchical Support Vector Machine Multi-class Classifier 1 1 2 ZHAO Hui , RONG Li-li , LI Xiao ( 1. Institute of System Engineering, Dalian University of Technology, Dalian Liaoning 116024, China; 2 . Xinjiang Institute of Physics Chemistry Technology, Chinese Academy of Sciences, Urumoqi Xinjiang 830011, China) Abstract: Designing the hierarchical structure is a key issue for the hierarchical support vector machine multi-class classifica- tion. Inter-class separability is an important basis for designing the hierarchical structure. A new method based on linear sup- port vector machines is proposed to measure inter-class separability. Furthermore, a method is presented which designs a hier- archical support vector machine multi-class classifier based on the inter-class separability. Experimental results indicate that the new method speeds up the hierarchical support vector machine multi-class classifiers and yields higher precision. Key words: Support Vector Machines; Multi-class Classification; Hierarchical Structure; Inter-class Se