并行采样Mapreduce移动社交网络k均值分布层次聚类  

MDHC: Parallel Sample Mapreduce Mobile Social Network k Means Distributed Hierarchical Clustering Analysis

作　　者：黄健恒[1] 于军琪[1] 赵敏华[1] 

机构地区：[1]西安建筑科技大学信息与控制工程学院,陕西西安710055

出　　处：《小型微型计算机系统》2016年第8期1780-1785,共6页Journal of Chinese Computer Systems

基　　金：住房和城乡建设部项目(2012-K1-1)资助;陕西省留学人员科技活动择优项目(陕外专发[2013]27号)资助;西安市城乡建设委员会基金项目(市建发2013)资助;中国建筑学会2014年立项学术研究课题(建会秘[2015]16号)资助

摘　　要：针对移动社交网络社区管理问题,研究提出聚类分析框架,并设计相应的节点信息交换数据结构,该结构能够存储节点相遇过程中的节点间相似度信息,并适用层次聚类进行分析.但存在数据处理量随节点数量增加而大幅增加的问题,对此为解决大型移动社交网络数据维数灾难问题,利用采样构建数据子集方式,并结合Mapreduce计算模型实现并行移动社交网络均值聚类中心选取方式.同时,为提高算法稳定性和便于设计信息转发方案,采用一种分布式层次聚类方式,对个体进行聚类分析和基于层级的转发方案设计.最后,通过实验,验证了所提算法在保持聚类精度前提下,有效提高了大型移动社交网络聚类分析效率.According to the problem of community management of mobile social network, here proposed the cluster analysis framework, and designed the corresponding node information exchange data structure, which could store the similarity information between nodes nearby, and meet the k-means clustering analysis. But the amount of data processing would be substantial increase with the increase of the number of nodes, in order to solve the dimension disaster problem of large data mobile social network, here used the data sampling method to construct the subset of the data, and realized the parallel mobile social network k-means clustering analysis combined with Mapreduce cal-culation model. At the same time, in order to improve the stability of the algorithm, here used a distributed cluster-ing method. Finally, through the experiment, the proposed algorithm effectively improved the efficiency of cluster-ing analysis of large mobile social network, without expensing the clustering accuracy.

关 键 词：并行Mapreduce 移动社交网络 均值 大数据 层次聚类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]