基于密度峰值的标签传播算法
Label Propagation Algorithm Based on Peak Density
吴卫江 1王星豪 1潘雪玲 1郑艺峰 2郑猋1
作者信息
- 1. 中国石油大学(北京)石油数据挖掘北京市重点实验室 北京 102249;中国石油大学(北京)信息科学与信息工程学院 北京 102249
- 2. 闽南师范大学数据科学与智能应用福建省高等学校重点实验室 漳州 363000;闽南师范大学计算机学院 漳州 363000
- 折叠
摘要
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.
Abstract
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.
关键词
密度峰值聚类/标签传播/节点权重/社交网络Key words
density peak clustering/label propagation/node weigh/social networks引用本文复制引用
基金项目
国家自然科学基金(61701213)
福建省自然科学基金(2019J01748)
福建省教育厅中青年项目(JAT190392)
出版年
2024
国家科技期刊平台
NSTL
万方数据