高效的信任机制线性文本分类方法  

High efficient linear text categorization model based on trust mechanism design

作　　者：焦庆争[1,2] 蔚承建[1] 

机构地区：[1]南京工业大学信息科学与工程学院,江苏南京210009 [2]安徽师范大学信息管理中心,安徽芜湖241000

出　　处：《计算机工程与设计》2010年第10期2367-2371,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(60703071);安徽省高校省级自然科学研究重点基金项目(KJ2009A63)

摘　　要：基于信任机制设计了一种无须特征选择的高效的线性文本分类方法。面向特征与文档类的信任关系,使用bata概率密度函数评估特征的可靠度,提出特征对文档类的忠诚度的计算模型,基于忠诚度实现简单的线性文本分类器。采用20Newsgroup、复旦中文分类语料、SEWM2007评测语料等3个具有典型特征的单标签语料集,以朴素贝叶斯、KNN为比照算法进行了比较实验。实验结果表明,相对于传统算法,该算法分类性能显著提高,对不均匀语料和高维特征处理表现出很强的稳定性,同时算法执行速度快,适于大规模文本分类。A text categorization approach based on trust mechanism design is discussed,which is linear as well as no need of feature se-lection.The trust relation between feature and document class is analyzed.By using beta probability density function feature reliability,calculation model of loyalty degree is presented.Furthermore,linear text classifier is realized based on loyalty degree.In the experiments,Naive Bayes and KNN are selected as two comparison classifiers,and 20Newsgroup,Fudan Chinese evaluation data collection and SEWM2007 research corpus data are used to evaluate the effectiveness of the techniques proposed.The experimental results show the method could improve significantly the performance for text categorization,is suitable for large-scale text categorization and contributes a good solution for the difficulties of text categorization,such as high dimension characteristic of feature,in homogeneity of corpus and execution efficiency.

关 键 词：文本分类 信任机制 可靠度 bata概率密度函数 自然语言处理 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]