摘要
贝叶斯算法因其简单、快速、分类精确度高等优点被广泛应用于垃圾邮件过滤中,然而随着时间的推移,概念漂移现象导致贝叶斯分类器准确率下降。针对此问题,提出了基于用户反馈的客户端贝叶斯动态学习算法,可自动学习新的邮件样本,计算复杂度较低。实验表明该方法能较好地适应概念漂移,满足邮件分类的个性化需求,有很好的实用性。
Nave Bayesian algorithm has been widely used in spam filtering for its simplicity,efficiency and accuracy;however,the concept drift has led to a decline in accuracy of Bayesian classifier as time goes on.To solve this problem,a client-side dynamic Bayesian learning algorithm based on users feedback is put forward.It can learn new samples automatically,and has low computational complexity.The experiments show that the method is well adapted to concept drift,can meet the requirement of personal e-mail classification,and has good practicability.
出处
《计算机应用与软件》
CSCD
2011年第9期116-118,共3页
Computer Applications and Software
基金
甘肃省自然科学基金项目(096RJZA084)
甘肃省教育厅研究生导师科研计划项目(0914-02
0814-4)
关键词
贝叶斯邮件过滤
概念漂移
动态学习
Bayesian spam filtering Concept drift Dynamic learning