摘要
针对传统社区发现技术已无法有效处理大规模移动社交网络数据的问题,基于图论知识、网络性质提出山地模型,设计了滑坡算法,采用GraphX分布式图计算框架实现了并行化社区发现算法。模型首先利用模块度的聚类思想初始化山地模型并求取网络中每条边的权重,然后利用滑坡算法,迭代削弱社区间的关系,最终获得网络的社区结构。大量真实和模拟移动社交网络数据上的实验结果表明:所提算法能解决传统社区发现算法无法处理的大规模网络社区划分问题,在保证具有较高的社区识别准确率前提下,在时间性能上较现有并行社区发现算法提高140%以上,16台服务器构成的集群对于1000万个结点构成的大规模网络进行社区发现的时间不超过10 min。
As the cardinality of the networks grows rapidly,traditional community detection techniques cannot handle large-scale data of mobile social networks in an effective fashion.Based on the graph theory,network properties,a mountain model is proposed,and a landslide algorithm is designed.The parallelization of the community detection algorithm is implemented by using the GraphX distribution graph computing framework.Firstly,the model initializes the mountain model via modularity clustering and computes the weight of each edge in the network.Then,the landslide algorithm is applied to iteratively weaken the relationships between communities.Lastly,the community structure of the network is obtained.The extensive experiments were conducted on a large number of real and simulated data from mobile social networks and the results show that the proposed algorithm can effectively cope with the problem of large-scale community partition which cannot be handled by traditional community detection algorithms.With the guarantee of high community recognition accuracy,the time performance can be improved by more than 140%when compared to other parallel community detection algorithms and it takes less than ten minutes for a cluster with sixteen servers to identity communities in large-scale networks having ten million nodes.
作者
韩楠
乔少杰
元昌安
黄萍
魏军林
彭京
周凯
HAN Nan;QIAO Shaojie;YUAN Chang'an;HUANG Ping;WEI Junlin;PENG Jing;ZHOU Kai(School of Management,Chengdu University of Information Technology,Chengdu 610103,China;School of Software Engineering,Chengdu University of Information Technology,Chengdu 610225,China;Software Automatic Generation and Intelligent Service Key Laboratory of Sichuan Province,Chengdu University of Information Technology,Chengdu 610225,China;Guangxi College of Education,Nanning 530023,China;Sichuan Provincial Department of Public Security,Chengdu 610014,China)
出处
《重庆理工大学学报（自然科学）》
CAS
北大核心
2020年第1期94-102,共9页
Journal of Chongqing University of Technology：Natural Science
基金
四川省科技计划项目(2018JY0448,2019YFG0106,2019YFS0067)
国家自然科学基金项目(61802035,61772091,71701026)
四川高校科研创新团队建设计划(18TD0027)
广西自然科学基金项目(2018GXNSFDA138005)
成都信息工程大学中青年学术带头人科研基金项目(J201701)
成都信息工程大学科研基金资助项目(KYTZ201715,KYTZ201750)
关键词
移动社交网络
社区发现
分布式计算
滑坡算法
模块度
mobile social networks
community detection
distributed computing
landslide algorithm
modularity
作者简介
韩楠,女,博士,副教授,主要从事移动社交网络与社区发现研究;通讯作者:乔少杰,男,博士,教授,主要从事人工智能与社交网络研究,E-mail:sjqiao@cuit.edu.cn。
