基于有监督Bayesian网络的垃圾邮件过滤

对影响邮件特性的邮件报文格式作了仔细的分析并对垃圾邮件的特征进行了分类归纳，在此基础上构建了一个有监督的Bayesian邮件分类网络。通过对该网络作Bayesian参数估计，实现了判定邮件类别的不确定推理。对不同邮件测试集的在线学习试验结果表明，有监督Bayesian邮件分类网络能够有效地实现垃圾邮件的相对完备特征学习，改善邮件过滤的准确率。

Spam filtering algorithm based on supervised Bayesian parameter estimation

To improve the reliability and completeness of spam filtering, the E-mail message format was carefully analyzed, and the spam characteristics were generalized and classified. Based on these analysis, a supervised Bayesian network for E-mail classifer was constructed. Parameter estimation on this network realized an uncertain inference to identify E-mail's sort. On-line learning for different E-mail testing sets shows that such a classifying network can ensure the classification and filtering efficiently. It practically provides a viable solution by building a supervised Bayesian classifying network to execute relatively complete characteristics learning and improve the accuracy of E-mail filtering.