垃圾邮件过滤中特征选择方法研究  

Research on the feature selection method for spam filtering

作　　者：王军 史科[2] 王辉 

机构地区：[1]合肥工业大学信息与网络中心,安徽合肥230009 [2]安徽大学计算机科学与技术学院,安徽合肥230039

出　　处：《合肥工业大学学报（自然科学版）》2009年第12期1863-1866,共4页Journal of Hefei University of Technology：Natural Science

摘　　要：文章对垃圾邮件过滤中的特征选择问题进行了研究,引入"词共现模型"考虑词语之间的语义联系信息,和传统的信息增益特征选择方法结合表示邮件,采用神经网络方法对邮件进行分类得到垃圾邮件过滤器。实验表明,文章提出的将词共现对和信息增益结合的特征选择方法能够提高垃圾邮件过滤的精确度。Feature selection for spam filtering is researched in this paper. The word co-occurrence model is introduced to analyze the semantic relation between phrases. Features representing emails are selected by word co-occurrence and information gain. The neural network is used to classify emails and construct the spare filter. The experiments show that the precision of spare filtering is increased by feature selection which combines word co-occurrence and information gain.

关 键 词：垃圾邮件过滤 信息增益 词共现模型 神经网络 交叉覆盖算法 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]