欢迎来到天天文库
浏览记录
ID:3906633
大小:743.20 KB
页数:11页
时间:2017-11-25
《基于话题综合因子分析的语义社会网络社区发现算法_杨静》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、计算机研究与发展DOI:10.7544?issn1000-1239.2014.20120793JournalofComputerResearchandDevelopment51(3):559-569,2014基于话题综合因子分析的语义社会网络社区发现算法杨静1辛宇1谢志强21(哈尔滨工程大学计算机科学与技术学院哈尔滨150001)2(哈尔滨理工大学计算机科学与技术学院哈尔滨150080)(yangjing@hrbeu.edu.cn)SemanticsSocialNetworkCommunityD
2、etectionAlgorithmBasedonTopicComprehensiveFactorAnalysis1,XinYu1,andXieZhiqiang2YangJing1(CollegeofComputerScienceandTechnology,HarbinEngineeringUniversity,Harbin150001)2(CollegeofComputerScienceandTechnology,HarbinUniversityofScienceandTechnology,Ha
3、rbin150080)AbstractAimingattheproblemthatgeneralsocialnetworkcommunitydetectionalgorithmcouldonlydetectthecommunitywithsolerelationship,whichcouldntrepresentthesemanticssimilarityofrealsocialcommunityandcouldntresolvetheproblemofSSNcommunitydetecti
4、onwithmultiplesemantictopic,weproposeanSSN(semanticssocialnetwork)detectionalgorithmbasedontopiccomprehensivefactoranalysis.Thisalgorithmfirstlydefinesthemultivariatesemanticsinformationastopic,takesmultivariateTCF(topiccomprehensivefactor)asthemeasu
5、rementoftopics,andthedifferenceoftopicdensityaspolymerizationdirection,andestablishestheinitialcommunitystructure.Secondly,weestablishthecostfunctionwiththegoalofminimizingthesemanticssimilarityinsidethecommunitiesandmaximizingthesemanticssimilarityb
6、etweendifferentcommunities,whensomeboundarynodeschangethecommunity.Thirdly,theSAOP(simulatedannealingoptimizationpolicy)isestablishedbasedontheinitialcommunityandcostfunction,whichtakesthevalueofcostfunctionasparameterwhentheboundarynodeschange.Wecou
7、ldoptimizetheinitialcommunitystructuregloballyandachievethesemanticscommunitydetectionwithmultivariate.Finally,theeffectivenessofSSNcommunitydetectionisprovedbyaserialofsimulations.Keywordssemanticssocialnetwork;semanticscommunitydetection;multivaria
8、tesemanticstopic;factoranalysis;simulatedannealing摘要针对一般社会网络社区发现算法仅考虑各节点的邻接关系,所划分的社区仅为一元关系社区,不能代表社区成员的语义相似性且无法处理具有多元语义话题的语义社会网络社区发现问题,提出基于话题因子分析的语义社会网络社区发现算法.该算法将节点的多元信息抽象为话题,先以多元话题综合因子作为节点话题信息度量,以节点间的话题密度差异作为节点聚合方向,构建初始社区结构;再以最大化社区内部话题信息相似度和最小化社区外部话
此文档下载收益归作者所有