作者:杨海陆，赵鑫，陈晨，王莉莉
Authors:YANG Hailu,ZHAO Xin,CHEN Chen,WANG Lili摘要:社区是一种介于微观和宏观之间的节点聚合体，对深入理解社交用户的行为规律具有重要意义。传统的社区发现方法假设节点在网络中具有同等地位，忽略了节点影响力在社区形成中的影响和作用。针对这一问题，提出一种基于节点影响力扩张的社区发现方法。首先，基于蒙特卡洛近似计算节点的局部影响力。其次，提出一种新的离心率计算方法对桥接节点进行筛选，增强种子质量。最后，采用动态规划对种子进行扩张生成社区。实验结果表明：基于节点影响力扩张进行社区发现能够有效的挖掘粒度较小的社区结构，在模块度、D-Score等指标具有一定的性能优势。
Abstract:Community structure is a type of node aggregate that exists on both a micro and macro scale, and it is critical to fully comprehend the behavior and law of social network users.Traditional community detection approaches presume that all nodes in a network have the same status, neglecting the influence and function of node influence in community formation.To solve this problem, a community detection method based on node influence expansion is proposed. To begin with, the local influence of nodes is computed using the Monte Carlo approximation. Then, a new eccentricity calculation approach is provided to screen bridge nodes and improve seed quality. Finally, to finish the process of community detection, dynamic programming is employed for seed expansion.The results of the experiments reveal that community detection based on node influence expansion can effectively discover the community structure with small granularity and has certain performance advantages in modularity, D-score, and other indicators.

PDF全文下载地址:

可免费Download/下载PDF全文