資源描述：
《基于局部語義聚類的語義重疊社區(qū)發(fā)現(xiàn)算法-論文.pdf》由會員上傳分享，免費(fèi)在線閱讀，更多相關(guān)內(nèi)容在行業(yè)資料-天天文庫。
1、計算機(jī)研究與發(fā)展DOI：10．7544ssn1OOO一1239．2015．20140308JournalofComputerResearchandDevelopment52(7)：I510—1521，2015基于局部語義聚類的語義重疊社區(qū)發(fā)現(xiàn)算法辛宇楊靜湯楚蘅葛斯喬(哈爾濱工程大學(xué)計算機(jī)科學(xué)與技術(shù)學(xué)院哈爾濱150001)(哈爾濱工業(yè)大學(xué)電氣工程及自動化學(xué)院哈爾濱150001)(xinyu@hrbeu．edu．cn)AnOverlappingSemanticCommunityDetectionAlgorithmBasedonLocalSemanticClusterXinYu，YangJing
2、，TangChuheng，andGeSiqiao(CollegeofComputerScienceandTechnology，HarbinEngineeringUniversity，Harbin150001)(SchoolofElectricalEngineeringandAutomation，HarbinInstituteofTechnology，Harbin150001)AbstractSincethesemanticsocialnetwork(SSN)iSanewkindofcomplexnetworks，thetraditionalcommunitydetectionalgorit
3、hmsdependingontheadjacencyinsocialnetworkarenotefficientintheSSN．Tosolvethisproblem，anoverlappingcommunitystructuredetectingmethodonsemanticsocialnetworksisproposedbasedonthelocalsemanticcluster(LSC)．Firstly，thealgorithmutilizestheGibbssamplingmethodtoestablishthequantizationmappingbywhichtheseman
4、ticinformationinnodesiSchangedintothesemanticspace，withthelatentDirichletallocation(LDA)asthesemanticmode1；Secondly，thealgorithmestablishesthesimilaritymatrixofSSN，withtherelativeentropyofsemanticcoordinateasthemeasurementofsimilaritybetweennodes；Thirdly，accordingtothecharacteroflocalsmall—worldin
5、socialnetwork。thealgorithmproposestheS-fitnessmodelwhichisthelocalcommunitystructureofSSN，andestablishestheLSCmethodbytheS—fitnessmode1；Finally，thealgorithmproposesthesemanticmode1bywhichthecommunitystructureofSSNismeasured，andtheefficiencyandfeasibilityofthealgorithmandthesemanticmodularityarever
6、ifiedbyexperimentalanalysis．Keywordssemanticsocialnetwork(SSN)；overlappingcommunitystructuredetection；latentDirichletallocation(LDA)；relativeentropy；Gibbssampling；localsemanticcluster(LSC)摘要語義社會網(wǎng)絡(luò)是一種包含信息節(jié)點及社會關(guān)系構(gòu)成的新型復(fù)雜網(wǎng)絡(luò)，因此以節(jié)點鄰接關(guān)系為挖掘?qū)ο蟮膫鹘y(tǒng)社會網(wǎng)絡(luò)社區(qū)發(fā)現(xiàn)算法無法有效處理語義社會網(wǎng)絡(luò)重疊社區(qū)發(fā)現(xiàn)問題．針對這一問題，提出基于局部語義聚類的語義社會網(wǎng)絡(luò)重疊社區(qū)發(fā)現(xiàn)
7、算法，該算法：1)以LDA(1atentDirichletallocation)模型為語義信息模型，利用Gibbs取樣法建立節(jié)點語義信息到語義空間的量化映射；2)以節(jié)點間語義坐標(biāo)的相對熵作為節(jié)點語義相似度的度量，建立節(jié)點相似度矩陣；3)根據(jù)社會網(wǎng)絡(luò)的局部小世界特性，提出語義社會網(wǎng)絡(luò)的局部社區(qū)結(jié)構(gòu)S—fitness模型，并根據(jù)S—fitness模型建立了局部語義聚類算法(1oca1semanticclusterm，LSC)
