应用非负值矩阵分解模型的社区挖掘方法

发布时间：2018-05-03 00:21

本文选题：社交网络 + 社区挖掘　；参考：《小型微型计算机系统》2014年06期

【摘要】：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.
[Abstract]:Online social networks are growing rapidly, and it is important to mine them to understand the characteristics of network structure. A community mining method based on non-negative matrix decomposition is proposed, which can decompose the social network matrix into matrix combinations suitable for discovering the relationship between users and communities and overlapping relationships between communities. The iterative updating rule is used to solve the decomposition matrix optimally, and the convergence of the updating rule is proved. In addition, aiming at the scale-free characteristic of social network, by using the attribute information of user nodes to calculate the similarity between users, and to establish an implicit connection to a large number of isolated users, the isolated users can be brought into the unified mining framework for community division. Thus solving the problem that isolated users can not accurately divide the community. Relevant experiments and practical applications show that this method can effectively mine online social networks.
【作者单位】：仲恺农业工程学院信息科学与技术学院;广东省服务计算工程技术研究中心;华南师范大学计算机学院;
【基金】：国家自然科学基金项目(61272067)资助国家科技支撑计划项目(SQ2011GX07E01500)资助广东省自然科学基金团队研究项目(S2012030006242)资助广东省重大科技专项计划项目(2012A080104019)资助广东省高校优秀青年创新人才培养计划项目(2012LYM_0077)资助
【分类号】：O151.21;TP393.09

【参考文献】