基于最小风险的单一表文件垃圾邮件过滤算法  

Filtering Algorithm of Single Table Document Junk Mail Based on the Minimum Risk

作　　者：张永花[1] 崔永君[1] 

机构地区：[1]兰州交通大学数理与软件工程学院,甘肃兰州730070

出　　处：《兰州交通大学学报》2009年第1期60-63,共4页Journal of Lanzhou Jiaotong University

基　　金：甘肃省自然科学基金(0803RJZA023)

摘　　要：研究探讨了传统贝叶斯模型的原理和优缺点,指出朴素贝叶斯算法没有考虑到合法邮件和垃圾邮件被误判带来的不同损失,因而有一定的局限性,而后以朴素贝叶斯为基础,结合最小风险的贝叶斯邮件过滤算法,提出改进的基于垃圾单词的单一表文件垃圾邮件过滤算法,提高了系统的执行效率,减少合法邮件的误判对用户造成的损害,并构建了邮件过滤模型.The limitation of Naive Bayes algorithm used for filtering mails is analysed. The traditional Naive Bayes method doesn＇t consider the different features between the legitimate mail and the junk in the classification process. On the basis of Naive Bayes algorithm and the Bayes of the minimum risk, a revised E-mail filter is introduced using only spammy words to improve efficiency of the system and reduce the misjudgment of legitimate E-mail,and build a model of E-mail filtering.

关 键 词：反垃圾邮件 贝叶斯 过滤模型 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]