期刊名称：International Journal of Computer Science and Network Security
印刷版ISSN：1738-7906
出版年度：2018
卷号：18
期号：5
页码：84-91
出版社：International Journal of Computer Science and Network Security
摘要：Networks can represent a wide range of complex systems such as social, biological and technological systems. In such complex environments, there are many challenges and problems that can be studied and considered. One of the most important issues in such systems that has attracted a lot of interests in recent years, is link prediction. Many studies have been accomplished on link prediction over the last few years, but the existing approaches are not satisfactory in processing topological information as they have high time complexity. Many researches in traditional methods assume that endpoint influence represented by endpoint degree, prefers to facilitate the connection between big-degree endpoints. In this paper, we propose a new link prediction approach using Louvain community detection algorithm [25] for clustering nodes in different groups and then estimating future links by precise analyzing of the nodes relationships. The experimental results demonstrate that our proposed methods outperform the base methods. The comparison analysis with main stream baselines on 10 benchmark datasets shows that the results have been effectively improved on link prediction accuracy.
关键词：Link prediction; community detection; Social networks; social network security
