【作者】 陈琳； 黄杰； 龚正虎；

【机构】 国防科技大学计算机学院；

【摘要】 <正>1引言随着Internet的快速发展,电子邮件已经成为人们生活中重要的组成部分。然而大量的商用垃圾邮件不断充斥着用户邮箱,对用户产生了极大的干扰。通过利用先进的计算机技术、人工智能技术对邮件采用自动分类方法已经成为了一种有效手段和必然趋势。它不仅可以实现方便快捷的分类效果,节省更多还原

【Abstract】 With development of Internet,Emails become an important part in people’s life.But junk mails of business flood mailbox constantly,which disturbs users awfully.There are some methods handle junk mails automatically, which learn from training samples.These methods can recognize Spam effectively under noise free condition. But when there are noises in training samples,the effects of classifiers are poor.Support Vector Machines have many native features that adapt to classify text.When applying Support Vector Machines(SVM)to Email classification, the paper puts forward an anti-noise mail classifying method.It has two characteristics:1)It selects the most important mail characteristics as aspects of multi-dimension space from candidates by statistics theory.2)It separates noise samples by preprocessing and builds mail classifier which is trained from noise free samples.Improved method based on SVM is proved achieving better effect.Compared with original SVM method and na(i|¨)ve Bayes method,classification precision of anti-noise SVM method is higher.更多还原