粗LVQ神经网络的垃圾邮件过滤

Spam filtering based on rough-LVQ network

结合粗糙集理论和LVQ神经网络的优点,提出了粗LVQ神经网络的垃圾邮件过滤模型。利用属性重要性启发式约简算法对邮件特征项进行约简,再利用LVQ网络模式分类能力进行垃圾邮件过滤。实验结果显示,提出的模型与单纯的粗糙集和LVQ网络相比,具有容错和抗干扰能力,减小了神经网络结构的复杂性,缩短了训练时间,提高了垃圾邮件过滤的准确率。

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.