一种基于潜在引用网络的专利价值评估方法.pdfVIP

计算机 研 究与发展 DOI：10．7544 ssnlOOO一1239．2015 JournalofComputerResearchandDevelopment 52(3)：649—660，2O15 一 种基于潜在 引用 网络的专利价值评估方法 冯 岭 彭智勇 刘 斌 车敦仁。 (软件工程国家重点实验室 (武汉大学) 武汉 430072) (武汉大学计算机学院 武汉 430072) 。(南伊利诺伊斯大学计算机科学系 美 国伊利诺伊州卡本戴尔 II62901) (peng@whu．edu．cn) A Latent—Citation—NetworkBasedPatentValueEvaluationM ethod FengLing～，PengZhiyong ，LiuBin～，andCheDunren。 (StateKeyLaboratoryofSoftwareEngineering，WuhanUniversity，Wuhan430072) (SchoolofComputer，WuhanUniversity，Wuhan430072) 。(DepartmentofComputerScience，SouthernIllinoisUniversity，Carbondale，Illinois，USA L62901) Abstract Patentvaluereferstotheexchangevalueofapatentinthepurchaseandtransaction．Itcan providepreciousinformation forthedecisionmakingsofpatentownersand buyers．Existing patent valueevaluation methodsaremainly based on training orcitation analysis．However，themethods basedon trainingdependtoomuchontheexperimentalparameters，whichresultsinweakcredibility． Ontheotherhand，themethodsbased on citation analysisonlyconsiderdirectcitationsduring the evaluationleaving indirectcitationsand novelty ofpatentneglected．Forthesereasons，thispaper Dresentsalatent—citation—networkbasedpatentvalueevaluationmethodtoevaluatethevalueofeach patentin which directcitations．indirectcitationsand novelty are allconsidered．First，thelatent citationassociationisdiscovered utilizing similaritybetween patentsandalatentcitationnetwork is established．Then，abasicalgorithm isimplementedtoeffectivelyevaluatethevalueofeachpatenton thenetwork．Further，an improved algorithm isproposedto solvetheproblem ofinefficiency ofthe basicalgorithm．Finally，to handle the value variations caused by the arrivalofa new patent，a dynamicpatentvalueevaluation algorithm isdesigned to efficiently updatethevalueoftheoriginal patents．Asshownintheexperiments，theproposedmethodsinthispaperareeffective． Key words patent