摘要
在文本分类中k-NN分类方法简洁而有效,但在多类分类问题中,由于类的重叠和属性的不足导致训练文本和类边界出现不确定性,而传统k-NN分类方法无法处理这种不确定性.该文借助于几种经典的不确定性推理方法:DS证据理论、模糊集理论、模糊-粗糙集理论,来改进传统k-NN文本分类方法,实验表明基于不确定性推理的方法能够提高文本分类的精度和召回率.
The k-nearest neighbors( k-NN ) categorization method is simple and effective in text categorization. The uncertainty of training documents and classes border would appear in multi-class categorization, because of the overlapping of classes and the lack of features. But the conventional k-NN method is unsuitable to deal with this uncertainty. In this paper, the several classical uncertainty reasoning methods, such as DS evidence theory, fuzzy..set theory, fuzzy-rough set theory, is applied to modify the conventional k-NN categorization method. The experimental test shows that these methods based on uncertainty reasoning improve the efficiency of text categorization.
出处
《江西师范大学学报(自然科学版)》
CAS
北大核心
2007年第4期383-386,共4页
Journal of Jiangxi Normal University(Natural Science Edition)
基金
国家自然科学基金(60663007)资助项目