资源描述：
《基于异构网络的微博时序排名方法-研究》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。
1、华中科技大学硕士学位论文AbstractRankingisafundamentaltaskfornetworkanalysis,benefitingtofilterandfindingvaluableinformation,andbecomeoneofthehottopicsthattheInternetandacademicsconcerned.Theconventionalrankingsfocusedonthesinglefactorsofhomogeneousnetwork,namely,allnodestypeint
2、hehomogeneousnetworkaresame,thustherankingpossessthestrongdeterminacyofthefactors.Fortheheterogeneousnetworks,thatcomposedofmultipletypesofnodeandcomplexreliancestructures,thereisamutuallyinformationimbalancerelationshipbetweendifferenttypesofnodes.Henceweapplythein
3、formationflowpropagationtooptimizetherankingresults.Meanwhile,thevarationandnondeterminacyofthetimefactorcanaffecttheoptimizationofrankingresultsaswell,andcanintroduceerrororbiasintoextractingandminingthevalueableinformation.Inthispaper,weproposedahierarchicalrankin
4、gmodelonheterogeneousnetworkbasedonthetimefactors’varationandtheranking’snondeterminacyofhomogeenousnework.Rankingonweiboforinstance,wemakefulluseofthediverserankinginhomogeneousnetworktoinitiatethewebpages,weiboandusersrankingresults,therebyobtainingthepre-rankingr
5、esultsoftheheterogeneousnetwork.Accordingtothetime-basedpropagationofweibo,weusethelogisticregressionmethodtofittheweibo'slifecyclecurve,andgainingthevaritionofthetimecharacteristics.Afterthatweoptmizetheweiboinitialrankbymeansoftheweibotemporalweight.Finally,combin
6、ingtherankingresultsofweb,webandusertobalancetheinformationbetweenheterogeneousnetworkbasedontheinfomationflowbetweendifferenttypeofthenode,andtoobtaintheoptimalrankingresultsconsequently.Hence,findingthevaluableandpopularweiboinformation.Theexperimentalteststakesin
7、aweiboforinstance,andwecrawl1.7millionusers’originalmicroblogdatathroughwebcrawlers.Accordingtotheexperimentalresults,wecanprovethattheTemporalHeteRankmethodcaneffectivelyimprovetheranking’saccuracy,real-timeandreliability.ThenweapplytheTemporalHeteRanktohotspotsdet
8、ectingapplication,andcomparewiththeexistinghotspotsdetectionmodel,thustoanalyzeandprovethefeasibilityofproposedweiborankingapproach.Keywor
