您好,欢迎访问三七文档
当前位置:首页 > 商业/管理/HR > 信息化管理 > 基于用户评论的信任预测方法研究
2014-12-122015-02-092011GXNSFA018156kx2012GDYCSZ2014691971-、ccgycai@gmail.com1988-1988-.*541004。。。。Epinions。TP181A1001-3695201604-1019-05doi10.3969/j.issn.1001-3695.2016.04.012Studyontrustpredictionbasedonuser’sreviewsCaiGuoyongWangLiyuanLyuRuiSchoolofComputerScience&EngineeringGuilinUniversityofElectronicGuilinGuangxi541004ChinaAbstractInonlinesocialnetworktrustisthebasisofreliableinteractionbetweenusersandthestrengthsofinteractional-soaffecttrustestablishment.Althoughmanyresearchershavestudiedapproachesoftrustmodelandpredictionmosttrustmodelingmethodsarebasedontheexistingtrustnetworkandlackthein-depthstudyofuserinteractionandcontents.Thispaperstudiedthesparsenessoftheoriginalnetworkandtheeffectofuser’sinteractionbehavior.Thereforeitpresentedano-veltrustpredictionframeworkbasedonbothatrustnetworkandtheinteractivecontextsbetweenusersandputforwardakindofmeasurementmechanismtoevaluatethestrengthofinteractionrelations.Combinedwiththeexistingtrustnetworklearnedandusedatrustpredictionthresholdvaluetopredictunknowntrustrelations.EmpiricalexperimentsconductesonEpinionsdatasetshowthattheunknowntrustrelationscanbeeffectivelypredictedcombiningwiththeuser’sinteractionbehaviorsandtheuser’sinteractionbehaviorscanimprovetheperformanceofthetrustpredictionmodel.Keywordsonlinesocialnetworktrustuser’sbehaviorrelationshipstrengthtrustprediction0Web2.0、、。1。Epinions、、。。2~1523EigenTrust、TidalTrust4。、、。“”、、。。33420164ApplicationResearchofComputersVol.33No.4Apr.2016。。15。。Liu7SVM。Kiyana8、、。Nikolay4。。9。。Guha10。。Borzymek11。Xiang12。Tang13。Zhang14。2Epinions.com。1Epinions。whotrustswhoistrus-tedbyWeboftrustproductratingreviewrating。、。1Epinions。2。2a。、。。。。、。b。G16WW=W1∪W2。2.1。c。WTbW1λ。d。abW2λ。2.1·0201·33。———G3。U=u1u2…umC=c1c2…cfI=I1I2…In。。uiIj∈cfuicfefiui→cfcfuipficf→ui。。。。Epinions5。。17。—wuiIj3wij=1uiIj3wij=0.5uiIjwij=0。uicfDfiuicfDfi=∑nj=1Ij∈cfwijdui1dui=∑|C|f=1∑|cf|j=1Ij∈cfwijuicf|C||cf|f。cfuiOficfuiOfi=∑nj=1Ij∈cfwijocf2ocf=∑mi=1∑nj=1j∈cfwijCfuii=123…mmn。reviewer。。uiuiHui=∑nj=1f_ratingIj/k3f_ratingIjuβIjkuβn。uiujSijuiuj4。Sij=∑|C|f=1Hui·Dfi·Ofj4Hiuβ|C|DfiOfjuβcf。4Wmn。1。1WRBH。W。//1RBU、C2uiki=0uihi=03Wij=04fori=0j=0tomn//5ifrij≥3then^rij=16elseif0<rij<3then^rij=0.57else^rij=08endfor//H9fori=0j=0f=1tomn|C|10ifexistuseruiinusersetU&&bjf==1//bif∈B11hi+=hijki++//hij∈H12hi=hi/ki//13endfor/*BbjfIjCf*/14fori=0j=0tomn15ifexistuseruiinusersetU//ui16dui=∑nj=1wij17forf=1to|C|18ifexistuseruiinusersetU&&bjf==1//bif∈01//uicf19wuicf=∑nj=1^rij20ifexistfieldcfinfieldsetC&&bjf==1//f21ocf=∑mi=1∑nj=1^rij22endfor23endfor//24fori=0j=0f=1tomm|C|25Wij1+=hj*wujcf/duj*wuicf/ocf26endfor2.2。W1。5。λ。2。rλλ。2λwWtW1。·1201·4λw。1Δ=0.012r0=03nt=04nd=05forλ=0to1stepΔ6fori=0j=0tomn7ifwij1≥λ&&wijt==18nt←nt+19ifwij1<λ&&wijt==110nd←nd+111endfor12rλ=nt/nt+nd13ifrλ>r0//rλr014r0←rλ15λw←λ16endif17endfor18returnλwrλ=∑x≥λnumxtnumtnumxtλnumt。。2。2。33.1Epinions。。。Epinions1。1Epinions#Users22165122165#Items2962771296277#Categories27127#Ratings3590161.5#Helpful35901616#TrustLink155323122165296277359016155323。1。1x%1-x%。x506065708090。18PA=|N∩P||N|NP|.|。。3.2TP、TP+Similar、Random。TP、、10TP+Similar11Random。x%x=506070809010%50%、60%、65%、70%、80%、90%4。4x。。。。2。2RandomTPTP+SimilarEpinions50%0.00270.12010.12760.130460%0.00250.13350.13570.141165%0.00240.14370.14810.154870%0.00170.14730.15020.168980%0.00150.15210.15570.184690%0.00160.15580.17020.20142。2TP、TP+Similar。Random。4。。、。·2201·33。。1LiuGuanfengYanWangOrgunMA.Socialcontext-awaretrustnetworkdiscoveryincomplexcontextualsocialnetworksC//ProcofNationalConferenceonArtificialIntelligence.2012101-107.2KamvarSDSchlosserMTGarcia-MolinaH.Theeigentrustalgo-rithmforreputationmanagementinP2PnetworksC//Procofthe12thInternationalConferenceonWorldWideWeb.NewYorkACMPress2003640-651.3UgurKGolbeckJ.Sunnyanewalgorithmfortrustinferenceinso-cialnetworksusingprobabilisticconfidencemodelsC//ProcofNa-tionalConferenceonArtificialIntelligence.20071377-1382.4NikolayKThomoA.Trustpredictionfromuser-itemratingsJ.SocialNetworkAnalysisandMining201333749-759.5WanitaSNepalSParisC.AsurveyoftrustinsocialnetworksJ.ACMComputingSurveys2013454Article47.6OhHKKimJWKimSW.Aprobability-basedtrustpredictionmodelusingtrust-messagepassingC//Procofthe22ndInternatio-nalConferenceonWorldWideWebCompanion.2013161-162.7LiuHaifengLimEPLauwHWetal.PredictingtrustsamongusersofonlinecommunitiesanepinionscasestudyC//Procofthe9thACMConferenceonElectronicCommerce.NewYorkACMPress2008310-319.8KiyanaZAghaieA.MiningtrustanddistrustrelationshipsinsocialWebapplicationsC//ProcofIEEEInternationalConferenceonIn-telligentComputerCommunicationandProcessing.S.l.IEEEPress201073-78.9WangDashunPedreschiDSongChaomingetal.HumanmobilitysocialtiesandlinkpredictionC//Procofthe17thACMSIGKDDInternationalConferenceonKnowledgeDiscoveryandDataMining.NewYorkACMPress20111100-1108.10GuhaRVKumarRRaghavanPetal.PropagationoftrustanddistrustC//Procofthe13thInternationalConferenceonWorldWideWeb.NewYorkACMPress2004403-412.11BorzymekPSydowMWierzbickiA.EnrichingtrustpredictionmodelinsocialnetworkwithuserratingsimilarityC//ProcofIEEEInternationalConfere
本文标题:基于用户评论的信任预测方法研究
链接地址:https://www.777doc.com/doc-4916538 .html