关键词: 粗糙集, 特征约简, LVQ神经网络, 邮件过滤, 网络安全

Abstract: Combining with the advantages of rough sets and LVQ network，this paper presents a spam filtering model based on rough-LVQ network.It reduces the feature items of emails by using heuristic algorithm for reduction based on the significance of attribute，and then filters spam by using the pattern classification capability of LVQ network.The results show that compared with the rough sets or LVQ network method respectively，the proposed model which is fault-tolerant and anti-jamming decreases the complexity of the structure of network，shortens the time of training and enhances the accuracy of filtering spam.

Key words: rough sets, feature reduction, LVQ-network, spam filter, network security