摘要： 在分析传统垃圾邮件过滤技术的基础上，提出了一种基于用户反馈的反垃圾邮件技术。该技术通过引入用户反馈机制，使用改进的朴素贝叶斯方法，构建面向特定用户的过滤器，从而进行垃圾邮件过滤。邮件语料库实验和原型系统的测试证明，该方法能够有效地降低误报率，提高反垃圾邮件系统的可用性，具有较好的实用效果。

关键词: 垃圾邮件过滤, 机器学习, 朴素贝叶斯方法, 用户反馈

Abstract: On the basis of analyzing the traditional spam filtering techniques, this paper presents a novel anti-spam technique based on the users feedback. By introducing the novel users’ feedback mechanism, the technique adopts an improved Naïve Bayesian approach to construct classifiers for specific users to fulfill spam filtering. Experiments on well-known mail corpus and prototype demonstrate that the technique is able to reduce false positives and improve the availability of anti-spam system.

Key words: Spam filtering, Machine learning, Naï, ve Bayesian approach, Users feedback