论文部分内容阅读
在线社交网络增长迅速,对其进行社区挖掘对于了解网络结构特征具有重要意义.提出一种基于非负值矩阵分解的社区挖掘方法,能够将社交网络矩阵分解为适合于发现用户与社区所属关系以及社区之间重叠关系的矩阵组合形式.该方法应用迭代更新规则对分解矩阵进行了优化求解,并证明了更新规则的收敛性.另外针对社交网络存在的无标度特性,通过利用用户节点属性信息计算用户间的相似性,对大量孤立用户建立隐式联系,可以将孤立用户纳入统一的挖掘框架进行社区划分,从而解决了孤立用户无法准确划分社区的问题.相关实验以及实际应用表明该方法可以有效对现实中的在线社交网络进行社区挖掘.
The rapid growth of online social networks and its community mining are of great significance for understanding the characteristics of network structure.This paper proposes a community mining method based on nonnegative matrix decomposition that can decompose the social network matrix into suitable for discovering the relationship between users and communities and Which is a matrix combination of overlapping relations.This method uses the iterative update rule to solve the decomposition matrix and proves the convergence of the update rule.In addition to the scale-free characteristic of the social network, this paper uses the user node attribute information Comparing the similarity between users and establishing implicit contact for a large number of isolated users, it is possible to include orphaned users into a unified mining framework for community partitioning so as to solve the problem that an isolated user can not accurately classify a community. The relevant experiments and practical applications show that the method can Effective community mining of real online social networks.