Abstract: Spam identification and filtering is one of the hot issues.And the rough set is a new data analysis tool to deal with ambiguity and uncertainty of knowledge；it has been successfully applied to many areas of classification.Combining rough sets with decision tree，a spam filtering solution based on rough sets and decision tree（RS-DT） was proposed.The feasibility of the solution was indicated by the experiments on the public email corpus.Comparison experiments were also made between SVM classifier，Bayes classifier and RS-DT model.The results show that the RS-DT model can not only reduce the error rate of judging the normal email as spam，but also improve adaptive learning of the filtration system.

Key words: spam, rough set, data mining, decision tree

摘要： 垃圾邮件的识别与过滤是目前研究的热点问题之一。而粗糙集是一种新的处理模糊和不确定性知识的数据分析工具，已被成功地应用到许多有关分类的领域。将粗糙集与决策树结合，提出一个基于RS-DT的邮件分类方案与模型，并进行了实验及结果分析。通过与朴素贝叶斯模型及SVM的比较，表明提出的基于RS-DT的模型可以降低把正常邮件错分为垃圾邮件的比率，提高过滤系统的自学习能力。

关键词: 垃圾邮件, 粗糙集, 数据挖掘, 决策树