Analysis of co-authorship network Case study of political Sciences from Microsoft Academic Graph (MAG)

Abstract

Withtheadvancementintechnology,networkshavegrownbiggerandevenbecom- ing moreandmorecomplex.Researchershavecontributedtheireffortsforidentifying differentstructuresofdifferentcomplexnetworks.Thereisaneedtodetectandiden- tify suchcomplexnetworkinordertoknowhowthesenetworksprovidecommunica- tion withinthem.Networkssuchassocialnetworks,neuronnetworks,emailnetworks and webhyperlinknetworksareknownascomplexnetworks.SocialNetworkAnalysis (SNA)fieldisfacilitatingandprovidingtechniquesandmethodstoexploreandanalyze such intricateorcomplexsystemsusingtechniquessuchasgraphtheories,network properties andcommunitydetectionalgorithms.Assortedcommunicationonsocial networksattractedresearchersattentioninrecenttime.Massivesharingofopinions, ideas, experiencesandexpertisehighlightcommunicationthroughsocialnetworks.So- cial networkshavebecomeaflourishingnetworkforsharingsuchinformation.Identi- fying andanalyzingthesecommunicationtrendshavegainedimportancefordetecting patterns amongpeopleinsocialnetworksusingcommunitydetectiontechniques. Co-authorship networkisalsobecomingcentralpointofattentiontomanyresearchers. Domain specificco-authorshipcommunitydetectionisevolvingareathatisemerging beside othercommunitydetectionidentificationpatterns.Inthisthesis,weperformed analysis onco-authorshipnetworkinthefieldofPublicRelationandPublicAdminis- tration usingcommoncentralitymeasures.TheauthorsbelongtothefieldofPublicAd- ministration andPublicRelationsrelatedtodifferentresearchandacademicinstitutes present fromallovertheworld.Wefoundcohesivegroupsofauthorsandweranked them onthebasisofcentralitymeasures.Wetookdifferentcentralitymeasuressuchas betweenness, degree,pagerank,andauthorityscore,closenessandeigenvector.Find- ings revealtheco-hesivenetworkofdomainspecificco-authorswhoarebestintheir knowledgesharingandcollaborationinthefieldofPublicRelationsandPublicAdmin- istration. Our case study analyze the patterns of collaboration in co-authorship network based onmodularityandcentralitymeasures.Inourproposedwork,wehaveusedmod- ularity andcentralitymeasuresfordetectingbestauthorsandco-authorswhobelongto variouscommunities.Wehaveproposedanauthor’scommunitydetectionframework that isbasedonmodularity.Ourproposedcommunitydetectionframeworkdetects author’scommunitywhichhavebestpublicationsandworksonnon-overlappingcom- munity structures.Theframeworkiscapableofidentifyingauthorcommunitiesthat havemanynumberofpublicationsandcitations.Wehavealsodevelopedstructured database corpustoevaluateourproposedcommunitydetectionapproach.Thisstruc- tured corpusisalsoavailableasastandardtest-bedtocomparedifferentcommunity detection approaches.