信息网络安全

信息网络安全 ›› 2022, Vol. 22 ›› Issue (3): 85-96.doi: 10.3969/j.issn.1671-1122.2022.03.010

• 理论研究 • 上一篇    下一篇

基于Affinity Propagation算法的半监督微博水军识别

林义钧1, 吴渝1(), 李红波2   

  1. 1. 重庆邮电大学网络空间安全与信息法学院,重庆 400065
    2. 重庆邮电大学创新创业学院,重庆 400065
  • 收稿日期:2021-08-18 出版日期:2022-03-10 发布日期:2022-03-28
  • 通讯作者: 吴渝 E-mail:Wuyu@cqupt.edu.cn
  • 作者简介:林义钧（1993—）,男,重庆,硕士,主要研究方向为数据挖掘|吴渝（1970—）,女,重庆,教授,博士,主要研究方向为网络舆情|李红波（1970—）,男,重庆,正高级工程师,硕士,主要研究方向为机器学习
  • 基金资助:
    国家自然科学基金(61903056);国家社会科学基金(17XFX013)

Internet Hirelings Semi-supervised Detection of Weibo Based on Affinity Propagation Algorithm

LIN Yijun1, WU Yu1(), LI Hongbo2   

  1. 1. School of Cyber Security and Information Law, Chongqing University of Posts and Telecommunications, Chongqing 400065, China
    2. School of Innovation and Enterpreneurship, Chongqing University of Posts and Telecommunications, Chongqing 400065, China
  • Received:2021-08-18 Online:2022-03-10 Published:2022-03-28
  • Contact: WU Yu E-mail:Wuyu@cqupt.edu.cn

RichHTML

15

PDF (PC)

81

摘要：

对微博网络空间中水军账户的识别研究,有助于清朗网络空间和维护社会安定。首先,文章针对微博水军不断进化、传统特征集无法覆盖现有水军特征,结合水军定义与原始特征,构造了新特征。然后,针对水军账户标注困难,无标注数据又没能充分利用的问题,提出了一种基于Affinity Propagation算法的半监督微博水军识别方法（APDHW）。该方法通过在Affinity Propagation算法中引入欧氏距离Radius阈值,再结合支持向量机分类算法,实现对微博水军识别。通过多组实验对比及实证研究,结果表明文章所提的微博水军识别方法在牺牲少量算法时间的情况下得到较好的识别效果,提升了水军识别的准确率和召回率。

关键词: 微博水军, Affinity Propagation, 半监督学习, 水军识别

Abstract:

The research on the Internet hirelings accounts in Weibo contributes to purify cyberspace and maintain social stability. First of all, in view of the continuous evolution of the Internet hirelings in Weibo, the traditional feature set cannot cover the existing features of it. Therefore, the new features are constructed combined with the definition of the Internet hirelings and its original features. Then, in view of the difficulty of account annotation and the insufficient utilization of no annotation data, a semi-supervised recognition method of the Internet hirelings in Weibo (APDHW) is proposed, based on Affinity Propagation arithmetic. In this method, the recognition of the Internet hirelings in Weibo can be implemented through bringing Euclidean distance Radius threshold in Affinity Propagation arithmetic and combining support vector machine classification arithmetic. Through a number of experiments and empirical research, the results show that the recognition method of the Internet hirelings in Weibo proposed in this paper achieves a better recognition effect under the expense of a small amount of arithmetic time, and improves the accuracy and recall rate of the Internet hirelings recognition.

Key words: Internet hirelings of Weibo, affinity propagation, semi-supervised learning, Internet hirelings detection

中图分类号: