基于多Bayes网的垃圾邮件智能过滤研究

A Way to Filter Junk Mail Intelligently Based on Multi-Bayesian Networks

在分析朴素 Bayes 方法用于垃圾邮件自动过滤中存在的一些问题基础上,提出了一种新的基于多 Bayes 网的垃圾邮件自动过滤方法。该方法利用多个 Bayes 网构成的多个分类器同时对邮件进行分类,当前邮件被认定是垃圾邮件当且仅当全部分类器都判断它为垃圾邮件。这种多个分类器同时工作及分类临界值的使用在一定程度上减少了将有用邮件误判为垃圾邮件的可能性。该方法还引入动态学习机制,在邮件分类过程中能够补充训练样本,满足不同用户的邮件分类标准。

Discussing some problems in filtering junk mail with Bayesian networks, a new way for filtering junk mail intelligently based on multi-Bayesian networks is proposed. Multi-classifiers are come from multi-Bayesian networks and mail is processed by Multi-classifiers. A mail is judged Spam if and only if every judging result is Spare. This idea and critical value can reduce the error probability of classifying. The adopted dynamic learning is introduced in the new method that is the multi-classifiers can be supplemented training examples so that satisfy different demand.