论文部分内容阅读
如何发现高质量的社区结构对于深刻研究和分析基于位置的社交网络(LBSN)这种新型复杂网络具有重要意义,然而,现有面向社交网络的社区发现方法都无法适用于具有多维异构关系的LBSN.为此,提出了一种基于联合聚类的用户社区发现方法 Multi-BVD,该方法先给出了融合用户社交网络与地理位置标签网络中多模实体及其异构关系的社区划分目标函数,然后使用拉格朗日乘子法得到目标函数极小值的迭代更新规则,并运用块值矩阵分解技术来确定最优的社区划分结果.仿真实验结果表明,Multi-BVD方法能有效地发现LBSN中具有地理特征的用户社区结构,该社区结构在社交关系和地理兴趣标签上都有更优的内聚性,并能更紧密地体现用户社区与地理标签簇间的兴趣关联性.
How to find a high-quality community structure is of great significance for the deep research and analysis of the new complex network based on location-based social network (LBSN). However, none of the existing social discovery methods for social networks are applicable to multi-dimensional heterogeneous relationships LBSN.Therefore, this paper proposes a multi-BVD method based on federated clustering for user community discovery, which firstly gives the target of community segmentation that combines multi-mode entities and their heterogeneous relations in social network and geo-tagging network Function, and then use the Lagrange multiplier method to obtain the iterative updating rule of the minimum value of the objective function, and use the block value matrix factorization technique to determine the optimal community partitioning result.The simulation results show that the Multi-BVD method can effectively Discovering the geographically characterized user community structure in LBSN that has better cohesion on social relationships and geo-interest tags and more closely reflects the interest relevancy between user communities and geo-tagged clusters.