基于密度峰值的标签传播算法

Label Propagation Algorithm Based on Peak Density

随着智能技术应用的推广,高质量社区的检测已成为社会网络研究的热点之一。由于具有线性时间复杂度,且无需预定义目标函数和社团数,标签传播算法(LPA)已得到广泛关注。然而,在标签传播过程中,LPA具有不确定性和随机性,进而影响检测社区结果的准确性和稳定性。为此,提出一种基于密度峰值的标签传播社区检测方法(DPC-RWL)。首先,采用密度峰值聚类算法查找出社区的核心节点集合,计算节点与核心节点集之间的权重,选取最大值为该节点赋予权值。最后,使用基于标签传播算法的归属度函数进行传播。真实网络和LFR人工基准网络的对比实验表明,所提算法能准确高效地识别出社区结构。

With the popularization of intelligent technology,high-quality community detection has become a hot topic in so-cial network research.Label propagation algorithm(LPA)has been widely attracted because of its linear time complexity and with-out predefining the objective function and community number.However,in label propagation,LPA is uncertainties and random-ness,which affects the group's accuracy and stability.Therefore,in this paper,a label propagation community detection approach based on peak density is proposed,called DPC-RWL.Firstly,the density peak clustering algorithm is employed to search the core node set of the community.Secondly,the weight between each node and the core set of nodes is calculated,and then the maximum value is selected as its weight.Eventually,the belonging degree function based on label propagation is utilized for propagation.The experiments between the real and LFR artificial benchmark networks show that the proposed algorithm can accurately and efficiently identify community structure.