基于Bayes参数估计的垃圾邮件过滤算法研究

朴素Bayes邮件过滤算法由于简单、易于理解,已被人们广泛接受,并应用到一些商用邮件系统当中.但面对目前垃圾邮件问题依然严重的现状,人们逐渐开始认识到采用简单的朴素Bayes邮件过滤算法已不能满足现有邮件过滤的性能要求.Bayes网络一直以来作为知识发现的一个重要分支,是人们研究的热点;邮件过滤问题也可以映射到一个Bayes决策网络模型中.通过构建针对邮件过滤的Bayes决策网络模型,并经过概率学习对关键节点作Bayes参数估计,可以实现邮件的概率分类发现.邮件样本试验结果表明新算法与朴素Bayes邮件过滤算法相比具有更快的收敛速度和更高的稳定性.

Research on Advanced Filtering Algorithm for Spam Email Based on Bayes Parameter Estimation

Presently, naive Bayes algorithm for Email filtering has been accepted widely for its simplicity and plainness. At the same time, the algorithm has also been applied in many commercial Email system However, facing with the persisted severe situation of spam Email, people are realizing gradually that depending on such a simple algorithm can' t meet the practical needs for anti-spam. On the other hand, Bayes network, as an important branch in knowledge dis- covery area, has been researched and explored for a long time. The question of Email filter can be projected to a Bayes network model too. By doing so and utilizing Bayes parameter estimation for key node in the model, we can accomplish Email classification and discrimination on probabilistic condition. According to the testing results, new algorithm has much higher rate to converge and stabilization than the naive Bayes one.