基于神经网络的电子邮件分类与过滤

E-mail classification and filtering based on neural network

现在电子邮件的应用非常广泛,已经成为人们生活中一种重要的通讯手段,但各种各样的垃圾邮件也是令我们十分困扰的问题,给出了一种电子邮件的分类过滤方法。电子邮件作为一种半结构化的文档,电子邮件信息包含了固定的语法部分和一定长度的可变文本部分,同时处理这两部分以得到更高的准确度。首先对邮件进行文本处理,得到特征向量;然后使用基于神经网络的方法对邮件进行分类过滤得到邮件分类器;最后通过实验验证分类器的有效性。

Now electronic mail （E-mail） is widely used, Which has become one of the most important forms ofcommtmication available, atthesametimejunkmailisquicklybecomingaseriousproblem. An E-mail classifying and filtering method is given. Modeled as semistructured documents, E-mail messages consist of a set of fields with predefined semantics and a number ofvariable length free-text fields. By dealing with both of them, higher accuracy is obtained. First, E-mail is dealt with and features vector are got. Then method based on neural network is used to classify and filter personal E-mails. The result showed that the classifier can perform filtering with reasonable accuracy.