A New SVM Based Method for Solving Multi-label Classification Problem
Multi-label classification problem is an extension of traditional multi-class classification problem in which its classes are not mutually exclusive and each sample may belong to several classes simultaneously.Such problems occur in many important applications (in bioinformatics,text categorization,scene classification,etc.).In this paper we propose a novel multi-label classification method based on adjusted probabilistie outputs of paired Support Vector Machine (SVM) classifiers for multi-label classification problem.In this method,one-versus-one decomposition technique is used firstly to divide a multi-label classification problem into many binary class sub-problems,in which some samples possibly are associated with two labels at the same time (overlapping area).Then we use two SVM classifiers to process each binary class sub-problem,and adjust probabilistic outputs of paired SVM classifiers to improve the classification accuracy on overlapping area.Experimental results on benchmark datasets Yeast and Scene show that our proposed method improves the classification accuracy efficiently,compared with some existed well-known methods.
Multi-label classification support vector machine probabilistic outputs of SVM
Benhui Chen Liangpeng Ma Jinglu Hu
Graduate School of Information,Production and Systems,Waseda University.2-7 Hibikino,Wakamatsu,Kitakyushu-shi,Fukuoka,808-0135,Japan
国际会议
云南大理
英文
325-334
2008-11-21（万方平台首次上网日期，不代表论文的发表时间）