微博中基于统计特征与双向投票的垃圾用户发现

发布时间：2018-06-28 19:46

本文选题：垃圾用户 + 信任传播　；参考：《计算机研究与发展》2013年11期

【摘要】：传统微博中垃圾用户发现主要依靠用户的显示统计特征.针对微博中关注网络的有向特性,给出了有向网络中局部三角形数量统计算法DirTriangleC,结合用户博文数量和局部三角形比例发现隐式垃圾用户;针对统计特征方法对垃圾用户误报和漏报的缺点,提出了基于统计特征与双向投票算法AttriBiVote,利用用户信任的双向传播与其邻居节点的统计特征共同决定用户类别.真实的Twitter数据集上验证了DirTriangleC和AttriBiVote算法的有效性,结果表明DirTriangleC算法能够发现约83.7%的"完全非活跃"状态的隐式垃圾用户,相对依靠显示统计特征方法增加了约2倍数量的疑似垃圾用户;同时AttriBiVote算法发现垃圾用户的数量和准确性均高于依靠统计特征的垃圾用户发现方法;最后实验分析了AttriBiVote算法的时间开销.
[Abstract]:In traditional Weibo, the discovery of garbage users mainly depends on the statistical features of the users. Aiming at the directed characteristic of Weibo, a local triangle statistic algorithm, DirTriangleC, is presented, which combines the number of user posts and the proportion of local triangles to discover the implicit spam user. In order to overcome the shortcomings of statistical feature method for false positives and omissions of garbage users, a bidirectional voting algorithm, AttriBiVote-based, is proposed, in which the bidirectional propagation of user trust and the statistical features of its neighbor nodes are used to determine the class of users. The validity of DirTriangleC and AttriBiVote algorithms is verified on real Twitter datasets. The results show that DirTriangleC algorithm can find about 83.7% of "completely inactive" implicit spam users. The number of suspected garbage users is increased by about 2 times, and the number and accuracy of garbage users by AttriBiVote algorithm is higher than that of garbage users who rely on statistical features. Finally, the time cost of AttriBiVote algorithm is analyzed.
【作者单位】：国防科学技术大学信息系统与管理学院;国防科学技术大学信息系统工程重点实验室;国防科学技术大学计算机学院;
【基金】：国家自然科学基金项目(60933005,91124002,61302144) 国家“八六三”高技术研究发展计划基金项目(2010AA012505,2011AA010702,2012AA01A401,2012AA01A402) 国家科技支撑计划基金项目(2012BAH38B04,2012BAH38B06) 国家“二四二”信息安全计划基金项目(2011A010) 国家“九七三”重点基础研究发展计划基金项目(2013CB329601,2013CB329601)
【分类号】：TP393.092

【参考文献】