基于带权超图的跨网络用户身份识别方法

发布时间：2018-01-18 14:45

本文关键词：基于带权超图的跨网络用户身份识别方法　出处：《计算机应用》2017年12期 　论文类型：期刊论文

【摘要】：随着各种社交网络的不断涌现,越来越多的研究者开始从多源的角度分析社交网络数据,多社交网络的数据融合依赖于跨网络用户身份识别。针对现有的基于好友关系(FRUI)算法对社交网络中的异质关系利用率不高的问题,提出了基于带权超图的跨网络用户身份识别(WHUI)算法。首先,通过在好友关系网络上构建带权超图来准确地描述同一网络中的好友关系及异质关系,以此提高表示节点所处拓扑环境的准确性;然后,在构建好的带权超图的基础上,根据节点所处拓扑环境在不同网络中大致相同这一特性,定义节点之间的跨网络相似性;最后,结合迭代匹配算法,每次选取跨网络相似性最高的用户对进行匹配,并加入双向认证和结果剪枝来保证识别准确率。在合作网络DBLP和真实社交网络上进行了实验,实验结果表明,在真实社交网络上,所提算法相比FRUI算法,平均准确率提高了5.5个百分点,平均召回率提高了3.4个百分点,平均F值提高了4.6个百分点。在只有网络拓扑信息的情况下,所提WHUI算法有效提高了实际应用中身份识别的准确率和召回率。
[Abstract]:With the continuous emergence of various social networks, more and more researchers begin to analyze social network data from the perspective of multiple sources. The data fusion of multi-social networks depends on cross-network user identification. Aiming at the problem that the existing friend relationship based FRUI-based algorithm is not high utilization of heterogeneous relationships in social networks. This paper proposes a cross-network user identification algorithm based on weighted hypergraph. Firstly, a weighted hypergraph is constructed to accurately describe the relationship and heterogeneity in the same network. In order to improve the accuracy of the topological environment in which the nodes are represented; Then, on the basis of constructing weighted hypergraph, the similarity of cross-network between nodes is defined according to the characteristic that the topology environment of nodes is roughly the same in different networks. Finally, combined with the iterative matching algorithm, the users with the highest similarity across the network are selected to match each time. Bidirectional authentication and result pruning are added to ensure recognition accuracy. Experiments are carried out on the cooperative network DBLP and the real social network. The experimental results show that it is on the real social network. Compared with the FRUI algorithm, the proposed algorithm improves the average accuracy by 5.5 percentage points and the average recall rate by 3.4 percentage points. The average F value is increased by 4.6 percentage points. In the case of only network topology information, the proposed WHUI algorithm can effectively improve the accuracy and recall rate of identity recognition in practical applications.
【作者单位】：国家数字交换系统工程技术研究中心;
【基金】：国家自然科学基金资助项目(61521003)~~
【分类号】：TP301.6
【正文快照】： 0引言多种多样的社交网络极大地丰富了人们的生活,人们通过QQ、微信与朋友保持联系,通过微博关注自己喜爱明星的动态,通过Linked In来发展职场社交。然而大多数社交网络间没有建立起公开的连接,因此用户的信息分散在多个社交网络中。识别出网民在不同网络中的虚拟账号的问题就

【相似文献】