电子学报 ›› 2016, Vol. 44 ›› Issue (9): 2062-2067.DOI: 10.3969/j.issn.0372-2112.2016.09.006

• 学术论文 • 上一篇    下一篇

基于社团结构的链接预测和属性推断联合解决方法

王锐, 吴玲玲, 石川, 吴斌   

  1. 北京邮电大学智能通信软件与多媒体北京市重点实验室, 北京 100876
  • 收稿日期:2015-02-05 修回日期:2015-05-27 出版日期:2016-09-25 发布日期:2016-09-25
  • 作者简介:王锐 女,1992年6月生于河南洛阳.北京邮电大学计算机学院硕士研究生.研究方向为数据挖掘与机器学习.E-mail:aboutstefanie@163.com;吴玲玲 女,1989年4月生于福建漳州.2015年毕业于北京邮电大学计算机学院.研究方向为数据挖掘与机器学习.E-mail:wulingling@bupt.edu.cn
  • 基金资助:

    国家重点基础研究发展计划(No.2013CB329602);国家自然科学基金(No.61375058,No.71231002);北京市高等教育青年英才项目(No.YETP0444)

Integrating Link Prediction and Attribute Inference Based on Community Structure

WANG Rui, WU Ling-ling, SHI Chuan, WU Bin   

  1. Beijing Key Lab of Intelligent Telecommunication Software and Multimedia, Beijing University of Posts and Telecommunications, Beijing 100876, China
  • Received:2015-02-05 Revised:2015-05-27 Online:2016-09-25 Published:2016-09-25

摘要:

链接预测与属性推断是社交网络数据挖掘的两项重要任务.之前的大部分研究工作将链接预测和属性推断视为不同的问题,分别研究解决方法.然而,根据网络结构的同质性理论,社交网络中的链接与属性之间具有内在关联.本文提出了基于社团结构的链接预测和属性推断联合解决方法(LAIC),将社团结构作为链接预测与属性推断的关联因子,利用用户属性和社团结构进行链接预测,利用链接信息得到社团属性进而推断用户属性.LAIC不仅同时解决了链接预测和属性推断问题,而且通过迭代使链接预测和属性推断的准确率可以相互提升.两个真实数据集上的实验证明LAIC方法是有效的.

关键词: 社交网络, 链接预测, 属性推断, 社团结构

Abstract:

Link prediction and attribute inference are two important tasks in social network mining.Most of the previous studies treated link prediction and attribute inference as different problems and sought for solutions separately.However,according to the theory of homophily,there are intrinsic relations between links and attributes in social network.We propose the link and attribute inference based on community (LAIC) solution which utilizes the community structure to connect link prediction and attribute inference.LAIC employs users' attribute and community structure for link prediction,and takes advantage of link information to get the attributes of communities for attribute inference of users.LAIC is not only able to predict attributes and links simultaneously,but also promotes the precision of link prediction and attribute inference mutually through iterations.Experiments on two real datasets verify the effectiveness of LAIC.

Key words: social network, link prediction, attribute inference, community structure

中图分类号: