文章翻译 1.docx

文章翻译 1.docx

《文章翻译 1.docx》由会员分享，可在线阅读，更多相关《文章翻译 1.docx（20页珍藏版）》请在冰豆网上搜索。

文章翻译1

ResearchonClusteringmethodBasedonAccessbehaviorofCustomerRequirements

BingLEI

SchoolofManagement,HenanUniversityofTechnology,Zhengzhou450001,China

Abstract:

Thispaperisbasedoncorporatewebsitevisiteddataandputforwardtheconceptofutilityofcustomerrequirementsalsobuildadatamodel.Onthisbasis,asimilarpreferenceofcustomersbasedoncustomerrequirementsclusteringmethodsisproposed.Thismethodincludescustomerrequirementutility（datamodeling）,clusterofcustomerrequirement（clusterofvisitors）,howtoconcludecharacteristicsofcustomersincluster（introductionofVisitorscharacteristics）.Amongthem,fordatamodeling,Adesignbasedonthe‘visitor-customerdemand’matrixofcustomerdemandutilitydatamodelismadeandputsforwardtwostepsconstructionmethodofconstructing.

Keyword:

accessbehavior、customerrequirement、datamodeling、cluster、corporatewebsite

1Foreword

Basedontheenterprisewebsiteoperation,resultinginalargenumberofaccesstorecords,Theseaccessrecordsareleftbythecustomersduringtheyvisitthesites,Thecustomer's‘footprint’containstheiraccessintent.Ingeneral,theenterprisewebsitecustomeraccessrecordsaremainlyclickstreamandproductreviewstwocategories.BasedontheComprehensiveinformationtheory[1]andconsumerinformationprocessinganddecisiontheory[2],basedontheseclickstreamandproductreviewswecaninferthatthecustomerdemandforpreference.Therefore，Basedontheinferredpreferenceofcustomerdemand,Customersareclassified,thiswillbepropitioustoenterprisetomakeacorrespondingmarketingplanningorproduceproductwhichcanmeetdifferentcustomerdemand.

BasedonWebsiteaccessrecordstoresearchthebehaviorofonlineconsumer,belongtothefieldofintelligentbusinessattheapplicationlevel,atthetechnicallevelintheminingareaofwebusage.ResearchonapplicationofWebusageminingintermsofconsumerbehaviormainlyconcentratedinthepersonalizedproductrecommendationsystemandknowledgeofbusinessinformationminingtwoterms.

Intheaspectofpersonalizedproductrecommendationsystem,mainlybasedonthebrosehistoryofproductonthewebsites,onthisbasistorecommendproductstocustomers.Atpresent,thistechnologyisrelativelymature,andhasappliedtoeachbige-commercesites,suchasA,etc.Fromtheperspectiveoftechnology,personalizedrecommendationtechnologymainlyincludesthecollaborativefilteringandcontent-basedrecommendationtwocategories.Intheaspectofthecollaborativefiltering,mainlyincludesmemory-basedrecommendation[3]andmodel-basedrecommendation[4]twoalgorithms.Intheaspectofcontent-based,thekeypointisthataccesstoinformation[5]andinformationfiltering[6],namelybyanalyzingtheintroductionofproductstorecommend.

Intheaspectofknowledgeofbusinessinformationmining,Buehner,etc.[7]discussedTheminingmethodofcommercialwebsiteknowledgeindetail,Bythismethod,miningattractcustomersFromtheWebdata,customerretention,crosssalesandotherbusinessintelligencerulescanbedone.;Komati[8],etc.discussedthemethodofanalysisofthelessonslearnedfromtheB2Cwebsites.Fromthemanagementperspectiveintheliterature[9]Confirmsthefactorsthataffectcustomertosearchforinformationinthenetwork;Literature[10]studiesthekeytechnologyofIntelligentBusinessSystemsbasedonWebusagemining;Literature[11]putsforwardanintelligente-commercemodel,onthisbasis,akindofcustomerPurchasingbehaviorExtractionalgorithmandakindofMulti-objectiveoptimizationmodelareproposed

Allroundtheexistingnetworkconsumerbehaviorresearch,Mainlyincludesconsumersearchbehavior，Commodity-relatedrules，Personalizedrecommendation,etc.Fewresearchonapplicationofenterprisewebsiteaccessrecordsascustomerdemandofthedatasource.

ThisarticleisbasedonenterpriseWebsitevisiteddataofaccessbehaviorofvisitorsandanalysistheutilityofcustomerdemand.Onthisbasis,asimilarpreferenceofcustomersbasedoncustomerneedsclusteringmethodsisproposed.Themethodmainlysolveshowtoconfirmcustomerdemandutilityaccordingtoenterprisewebsiteaccessbehavior,Customerdemandclustering,howtoconcludeinclustersofcharacteristicsofcustomersandotherissues.

2.Customerdemandclusteringmethod

Basedontheinformationtheory[1],accesstocustomerneedsfromcustomerrecords,essentiallyfromthegrammartopragmaticslayer"progressive",namelyfromthesyntaxlayerof"clickstream"andthecommenttext,tothetermcustomerdemandatthesemanticlevel,andthentothecustomersdemandcharacteristicsofpragmaticlayertransformation.Generallyspeaking,everyvisitortovisitthewebsite（1ormore）,producetheclickstreamandproductreviewdata,whichcontainsa"customerneeds"utility,therefore,toextractitfromthecustomerdemandsandutilityfunction,andthenestablish"visitors-customerdemandmatrix",onthisbasis,bythecustomerclustering,finallyhavethesame（orsimilar）needsofcustomers.

Figure1issuitableforenterprisecustomersclusteringmethodbasedonwebsiteaccessbehaviorframework.

Figure1customerneedsclusteringmethodbasedontheaccessbehaviorframework

Thebasicideaofthismethodisthatfirstofall,throughtheWebserverlogs,anddatabaseaccessto"clickstream"ofbusinessandthecommenttextdata,anddatamodeling.

Processdatamodelingisactuallyavisitoraccessrecordsfromthegrammartopragmaticslayer（customerdemandperspective）transformation.Inthetwosteps,oneisfromthesyntaxleveltosemanticlevel,twofromthesemanticlayertopragmaticlayer.

Theexpressionofsemanticlayer,thisarticleadoptsthewayofcustomerdemandcharacteristicsofkeywords.Herearetwokindsofcircumstances,forvisitorsofpageviews,willcombinewebsitecolumnsstructureextractiontopickeywordsonthepage;Fortheproductthecommenttext,willcommentonproductkeywordsextractedfeatures.Byextracting,itwillbebasedoncustomerdemandcharacteristicsofkeywordsemanticinformation.

Semanticlayer,however,getthekeywordswhilecanreflectthecharacteristicsoftheneedsofcustomers,butcannotexpresstheweightstothedifferentneedsofcustomers,alsocan'tclickontheflowandthecommenttextsemanticexpression,therefore,inpragmaticlevel,oneistheneedtoconstructtheutilityfunctiontofurtherclearcustomerdemand,thesecondistocombinetwoaccessrecordsofcustomerdemand,theresulting"visitors-customerrequirementsmatrix".

Datamodelingiscompleted,accordingtotherequirementof"customerclusteringvisitors-customerdemandmatrix",andsummarizethedemographiccharacteristicsofsimilardon'tcustomers,withthefinaldecisionsupportforenterprises.

3.Datamodeling

Needfordatapreprocessingbeforedatamodeling.Datapreprocessingincludingsurveyedpagerecognitionandcustomerrecognition,sessionidentificationandpathcompletion,customerdemandofkeyextract,etc.DatapreprocessingresultsgotacontainIinterviewedpage,jacustomerdemandofkeysetUC"visitors-surveyedcontent":

UC={UC1,UC2,…,UCk}，

Thereinto,

UCk=（UIDk,PVk,RKk）,KIdentifieroftheUIDkforvisitors.PVkkpageforvisitorsaccesstothecollection.RKkcustomerneedsinproductreviewsforvisitorsofkeyset.

PVk={（Pk1,Tk1）,（Pk2,Tk2）,…,（PkI,Tki）},PkIkforvisitorstoaccessthepageURL,TkIkforvisitorstoaccesstheIthelengthofthepage.

RKk={（Kk1,Nk1）,（Kk2,Nk2）…,（Kkj,Nkj）},AKKJKmentionedintheproductreviewsforvisitorstothejfeature,NK1asthenumberofvisitorstotheKJfeaturewords.

Accordingtothefigure1shows,Datamodelingisthekeypointofthismethod.Theessenceofwhichistobuildaccessbehaviorofthecustomerneedstheutilityfunction.Currently,aboutaccessbehaviorbasedutilityfunctionofthedemandintheproductresearchareveryfew,butalsoaresearchdifficulties.Therefore,asanexploratorystudy,thispaperpresentsamethodofbuildingutilityfunctionbasedoncustomerdemandcharacteristicsofkeywords,namelydatamodelingmethod.

Thebasicideaofthismethodistosimplifytheaccessbehaviorofcustomer’sneedsutilityforabinaryset,namely（keywords,weight）.ThisexpressionisalsotypicalbusinessdatamodelingmethodinWebusemining[12].

Isproposedinthispapertobuild"visitors-customerrequirements"matrixtorepresenttheneedsofcustomersutilitydatamodelingmethod,thedatamodelingprocessisdividedintotwosteps.AwasestablishedaccordingtotheUCvisitors-pagequestioned"thematrix,"visitors-key"andsecond,onthisbasis,thecombinationofcustomerdemandknowledgebase,buildcontentenhancedmatrix--"visitors-customerrequirements"matrix.

3.1"visitors-page"questionedthematrix.

Definition1"Visitors-pagequestionedmatrix"referstothevisitorsasaline,websitealltheneedanalysisofthepageascolumn,valuesastheaccesstime（withoutaccesstothevalue0）matrix.

AssumesthatthewebsitealltheneedanalysisofthesetofpagesforP,:

P={P（1,2,P...,Pm},andvisitorssetkofthepage,asaresult,visitthelengthineachrowinthedata,therearem-（I）avalueof0.

Indatapretreatmentphase,inordertopathcompletion,oftenaccordingtothefirstSession,Sessionidentificationissetbutavisitornumberofsessionsoveraperiodoftimetendtobemorethanone,andeachSessionofthesurveyedpagesmayrepeatvisit.Therefore,whenvisitorstogenerateksetofpage,needtomergethesamepage,itsaccesstimebyaccumula