基于社会计算和机器学习的垃圾邮件识别方法的研究  

Research on spam identification based on social computing and machine learning

作　　者：董源[1] 徐雅斌[1,2] 李卓[1,2] 李艳平[1] 

机构地区：[1]北京信息科技大学计算机学院,北京100101 [2]北京信息科技大学网络文化与数字传播北京市重点实验室,北京100101

出　　处：《山东大学学报（理学版）》2013年第7期72-78,共7页Journal of Shandong University(Natural Science)

基　　金：国家自然科学基金资助项目(60973107);网络文化与数字传播北京市重点实验室资助项目(ICDD201106);国家社会科学基金重大项目(12&ZD234);网络文化与数字传播北京市重点实验室开放课题(ICDD201207)

摘　　要：在对目前各种垃圾邮件识别方法进行研究分析的基础上,结合社会计算的理论和机器学习的方法,提出了一种新的垃圾邮件识别方法。通过利用邮件头部中能反映联系人社会关系的特征来构造一张联系人来往关系图对垃圾邮件进行初次识别,对于无法确定存在社会关系的联系人的邮件再利用机器学习的方法进行识别。实验结果表明,采用该方法进行垃圾邮件识别较之单纯采用贝叶斯方法,识别准确率有了较大的提高,同时,识别时间得到降低。Based on the investigation and analysis of the current various spam recognition methods, a new spare identifi- cation method is proposed inspiring by social computing theory and methods of machine learning. Firstly, initial recog- nition of spares is taken using a relationship map of the interactions among contacts, which is constructed with the help of the characteristics in the mail heads reflecting the social relation of contacts. After that, for the mails of the contacts which are not able to be identified having social relation, recognition methods based on machine learning are taken. Through the experiments, it is demonstrated that the proposed method can identify spares more accurately while taking a shorter time, comparing with the ones based on Naive Bayes.

关 键 词：社会计算 垃圾邮件识别 社会关系 机器学习 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]