基于贝叶斯神经网络的垃圾邮件过滤方法

垃圾邮件过滤是当前互联网应用中急需解决的一个重要课题，日益受到人们的关注。本文提出了一种基于贝叶斯神经网络BNN(Bayesian Neural Network)的垃圾邮件过滤方法，利用贝叶斯推理和神经网络相结合的贝叶斯神经网络算法对用户给定的正常／垃圾邮件集合进行训练，得到邮件过滤模型。并且提出了一种有效的特征选择方法，采用信息增益准则，有效降低了特征维数。经过实验测试，本文提出的方法可以实现对垃圾邮件的有效过滤。

A Spam Filtering Method Based on Bayesian Neural Network

Spam filtering is an important task in the application of Internet, which receives increasing emphasis. In this paper a method of spam filtering based on the Bayesian Neural Network (BNN) algorithm is presented. The Bayesian approach is used for neural network to learn from the user given training spam/normal e-mail set. And the number of features needed as the input of the BNN model is reduced effectively through the proposed feature selection approach, where information gain is chosen as a criterion. The result of the experiment shows that the method in this paper can filter spam effectively.