基于节点属性的重叠社区发现算法改进

摘要

复杂网络越来越多地应用到各个领域,而在复杂网络中存在着一定的社区结构,社区发现的目标就是从无标记的网络中发现具有相似功能的节点组合.分析基于社区隶属模型AGM的重叠社区发现算法的特点及机理,讨论节点属性对社区发现的可能影响,通过节点属性来调整节点对社区的隶属情况和节点之间生成边的概率,获得更合理的社区隶属指标和对应的社区划分.真实数据集上的实验表明,改进后的算法能够提高采样效率.%Complex network is increasingly applied to various fields. However, there exists certain community structure in complex network. The goal of community discovery is to find similar nodes with similar functions from the non-mark network. The overlapping community based on the community membership model AGM is analyzed, and the characteristics and mechanisms of the algorithm thus found, and the possible impact of node properties on community discovery also discussed. Through node attribute, the node's membership of the community and the probability of the generated edges of nodes are adjusted to get a more reasonable community membership index and the corresponding community partition. Experiments on real data sets indicate that the Modified algorithm could improve the sampling efficiency.