基于支持向量机的文本兼类标注

SVM-based Documents Multi-labeling

该文分析了现有多类别支持向量机分类器的特点及DAGSVM的优势,并结合模糊技术改造DAGSVM使之能进行兼类标注的多类别分类。改进后的FDAGSVM采用模糊决策面代替了DAGSVM的分明决策面,使判决过程适应兼类标注的要求,克服了传统的多类别分类支持向量机必然将样本分入某一类别的不足。基准数据的兼类标注多类别分类试验表明,FDAGSVM在文本的兼类标注分类中表现出较好的性能。

This paper analyses the characteristics of proposed multi-class SVMs and explains the superiority of DAGSVM over others.Based on DAGSVM and fuzzy technology,we propose the FDAGSVM,which can multi-label samples.Using the fuzzy decision hyperplane instead of the crisp decision hyperplane,FDAGSVM optimizes the decision process and overcomes the shortcoming of traditional MSVMs.Mulfi-labeling tests on benchmark data show that FDAGSVM performs well.