基于Winnow算法的垃圾邮件过滤器研究  

Study on spam filters based on Winnow algorithm

作　　者：陈峰[1] 

机构地区：[1]浙江省公安厅科技通信管理局,杭州310009

出　　处：《计算机应用》2009年第B12期21-23,26,共4页journal of Computer Applications

摘　　要：实现了基本的Winnow算法、Balanced Winnow算法和带反馈学习功能的Winnow算法,并将其成功地应用于大规模垃圾邮件过滤,分别在SEWM2007和SEWM2008数据集上对上述三个算法进行了对比实验。实验结果表明,Winnow算法及其变体在分类效果和效率上都优于Logiistic算法。In this paper, the basic Winnow algorithm, the balanced Winnow algorithm, and the improved Winnow algorithm with feedback learning function were implemented and applied in the domain of large-scale spam filtering. Comparison experiments were conducted on two datasets of SEWM2007 and SEWM2008 respectively. Experiment results show that the basic Winnow algorithm and its improved algorithms are more accurate and efficient than Logistic regression algorithm while classifying.

关 键 词：垃圾邮件过滤 WINNOW算法 LOGISTIC回归 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]