一种改进的多分类孪生支持向量机  

AN IMPROVED MULTI-CLASS CLASSIFICATION TWIN SUPPORT VECTOR MACHINE

作　　者：周开伟 钱雪忠[1] 周世兵[1] Zhou Kaiwei;Qian Xuezhong;Zhou Shibing(School of Internet of Things Engineering,Jiangnan University,Wuxi 214122,Jiangsu,China)

机构地区：[1]江南大学物联网工程学院,江苏无锡214122

出　　处：《计算机应用与软件》2022年第4期269-274,299,共7页Computer Applications and Software

基　　金：国家自然科学基金项目(61673193);中央高校基本科研业务费专项资金项目(JUSRP11235)。

摘　　要：现有的多分类孪生支持向量机主要考虑支持向量机的经验风险最小化原则,而忽略了结构化风险。针对该问题,通过引入正则项式,实现算法的结构风险最小化原则,并结合多对一的组合策略和最小二乘法,提出一种改进的最小二乘多分类孪生支持向量机。在UCI数据集上的实验研究表明,该算法相对于传统的多分类支持向量机在分类性能上有明显提高。The existing multi-classification twin support vector machines mainly consider the empirical risk minimization principle of support vector machines and ignore the structural risk. In order to solve this problem, an improved least-squares multiple classification twin support vector machine was proposed by introducing regular term expression, combining many-to-one combination strategy and least square method, which realized the structural risk minization principle. The experimental results on UCI dataset show that the proposed algorithm has better classification performance than traditional multi-classification support vector machines.

关 键 词：孪生支持向量机 多对一 多分类 最小二乘 结构风险最小化 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术] TP18