基于中心法的多层次文本分类方法的研究  

Research on a hierarchical text categorization method based on centroid

作　　者：何尧[1] 陈治平[1] 杨亚蕾[1] 

机构地区：[1]福建工程学院计算机系,福州350014

出　　处：《信息技术》2007年第12期116-118,共3页Information Technology

摘　　要：中心法运算速度快,效率高,而多层次分类器能有效地应对较多类别的分类任务,为此,提出了基于中心法的多层次分类法,通过分析大量类别之间的关系,把类别组织成树状结构,并在特征选择时,根据层次结构特色采取去根处理,在分类时采用中心法来进行分类。经过实验,与一般的层次分类算法、平面分类算法进行比较,该分类法具有较好的性能。Centroid-based methods is a high efficient class of methods for text categorization, and hierarchical classification can deal with the classification task of the many categories efficiently, So this paper presents a new approach that combines these two methods, through analyzing the relationship between many classis, organizing the categorize into a tree structure, reducing each word to its root to decrease the effect of word variations on the classification, and automatically classifies a document into one or more predefined categories using centroid-based method. Finally, the experiment results show that the new approach, proposed in this paper, outperforms the flat or generic hierarchical methods with improved accuracy.

关 键 词：中心法 多层次分类 文本分类 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]