Automated Detection of Influential Patents Using Singular Values

Abstract

Centralitymeasuressuchasdegreecentralityhavebeenutilizedtoidentifyinfluentialandimportantpatentsinacitationnetwork.However,noexistingcentralitymeasurestakeintoconsiderationinformationfromthechangeofthesimilaritymatrix.Thispaperpresentsanewcentralitymeasurebasedonthechangeofanodesimilaritymatrix.Theproposedapproachgivesmoreintuitiveunderstandingofthefindingoftheinfluentialnodes.Thepresentstudystartsoffwiththeassumptionthatthechangeofmatrixthatmayresultfromremovingagivennodewouldassesstheimportanceofthenodesinceeachnodemakeacontributiontothegivensimilaritymatrixbetweennodes.Thevariousmatrixnormsusingthesingularvaluessuchasnuclearnormwhichisthesumofallsingularvalues,areusedforcalculatingthecontributionofagivennodetoanodesimilaritymatrix.Inotherwords,wecanobtainthechangeofmatrixnormsforagivennodeafterwecalculatethesingularvaluesforthecaseofthenonexistenceandthecaseofexistenceofthenode.Then,thenoderesultinginthelargestchange(i.e.,decrease)ofmatrixnormscanbeconsideredasthemostimportantnode.Computationofsingularvaluescanbecomputationallyintensivewhenthesimilaritymatrixsizeislarge.Therefore,thesingularvalueupdatetechniqueisalsodevelopedforthecaseofthenetworkwithlargenodes.WecomparetheperformanceofourproposedapproachwithotherwidelyusedcentralitymeasuresusingU.S.patentsdataintheareaofinformationandsecurity.Experimentalresultsshowthatourproposedapproachiscompetitiveorevenperformsbettercomparedtoexistingapproaches.