基于用户行为的邮件分类算法

发布时间：2018-06-19 18:47

本文选题：分类算法 + 垃圾邮件　；参考：《计算机应用》2014年05期

【摘要】：针对垃圾邮件过滤过程中分类模型难以个性化、难以适应用户兴趣动态变化的问题,提出了一种基于用户行为的邮件分类算法。通过分析朴素贝叶(NB)斯分类算法的原理,改造朴素贝叶斯算法,使其具有动态调整能力。邮件服务器接收到新邮件后自动进行分类判别,用户浏览邮件的过程中对邮件进行操作,根据用户对错分邮件的处理自动将该邮件加入训练数据集,并动态更新相应特征的统计概率,使邮件分类算法能够依据用户对不同邮件的操作行为动态调整分类模型,以达到有效过滤垃圾邮件的目的。与常用的贝叶斯分类算法的实验比较表明在给定小样本集合进行训练的情况下,新算法对于垃圾邮件的识别率比传统的朴素贝叶斯方法、基于风险敏感的朴素贝叶斯方法等提高了10%,获得了较好的分类性能。
[Abstract]:In order to solve the problem that the classification model is difficult to personalize and adapt to the dynamic changes of users' interests in spam filtering, a classification algorithm based on user behavior is proposed. By analyzing the principle of NBs classification algorithm, the naive Bayes algorithm is modified to make it have the ability of dynamic adjustment. After receiving the new mail, the mail server automatically classifies and discriminates. During the process of browsing the mail, the mail is operated, and the mail is automatically added to the training data set according to the processing of the user's right and wrong mail. The statistical probability of the corresponding features is dynamically updated so that the mail classification algorithm can dynamically adjust the classification model according to the user's operation behavior to achieve the purpose of filtering spam effectively. The experimental results show that the new algorithm is more efficient than the traditional naive Bayesian algorithm in the case of training the set of small samples. The naive Bayes method based on risk sensitivity improves 10% and achieves better classification performance.
【作者单位】：长沙学院信息与计算科学系;
【分类号】：TP393.098

【参考文献】