改进的贝叶斯垃圾邮件过滤算法  

An improved Bayes algorithm for filtering spam e-mail

作　　者：王美珍[1] 李芝棠[1] 吴汉涛[1] 

机构地区：[1]华中科技大学计算机科学与技术学院,湖北武汉430074

出　　处：《华中科技大学学报（自然科学版）》2009年第8期27-30,共4页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(60573120);国家高技术研究发展计划资助项目(2007AA01Z420)

摘　　要：为了减少垃圾邮件误判造成的影响,在传统的贝叶斯算法基础上提出了相应的改进措施:引入损失因子来评估垃圾邮件误判件时带来的风险.通过理论推导和实验验证,得出损失因子的最佳值,来改善正常邮件的误判问题.最后通过实验测试和结果分析,表明基于改进的贝叶斯算法在垃圾邮件过滤中可以有效地减少误判,使查全率和查准率达到一个比较理想的效果.In order to reduce the influence of spam false negative result, an improment solution based on the traditional Bayesian algorithm is proposed, in which the loss factor is introdued to evaluate the risk of spare false negative rate. Through the theory inference and experimental verification, the best value of the loss factor was obtained, which can improve the misjudgment problem of the Bayesian algorithm. Lastly, by experiments testing and ianalysis, the result indicats that the improved Bayesian algorithm can reduce the false negative error rate when filtering spam e-mail, and take the recall and precision rates into correspondly ideal effect.

关 键 词：垃圾邮件 最小风险 贝叶斯 损失因子 误判 风险 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]