不同用户对垃圾邮件的判定有所差别,考虑到同一用户的自认垃圾邮件相似度较大,提出对特定用户进行针对性的垃圾邮件过滤方法.系统除重点利用邮件正文信息外,还尝试加入发件人、群发信息和主题相关度信息,改 进朴素贝叶斯公式用于邮件正文的概率计算,基于BP神经网络构造垃圾邮件判别系统.实验表明,改进的朴素贝叶斯公式用于本文的系统是可行的,基于BP神经网络的垃圾邮件过滤系统能有效综合以上四项数值进行全局判别,进而对特定用户的邮件产生不错的过滤效果.%Different users may have different judgment upon the spam. Considering that the spams judged by the same user may have higher similarity, a spam filtering method specially made for a specific user was presented. Besides the emphatic utilization of the e-mail text information, the sender's information, cluster-sending information and topic similarity information were also tried to join in the system. The Naive Bayes formula was improved to calculate the probability of the e-mail text. A spam judging system based on the BP neutral network was constructed, also. Experiment showed that it was feasible to use the improved Naive Bayes formula in the system presented. The spam filtering system based on BP neural network could effectively integrate the above-mentioned four informations and globally judge the spam bringing a better effect to specific user's e-mail.
展开▼
