基于RFID技术的智能服装推荐系统设计_王雪

DesignofClothingRecommendSystembasedonRadioFrequencyIdentification XueWANG1,AihuaDONG1,YizhiWU11CollegeofInformationSciencesandTechnology,DongHuaUniversity,Shanghai,ChinaEmail:[e-mail],[e-mail],[e-mail]:Byusingdatabasemanagementandautomaticrecommendationtechnology,aClothingRecommendSystembasedonradiofrequencyidentification(RFID)technologyinapparelretailingstoreisdesignedforthepurposeofimprovingtheclothingrecommendpertinence.Therecommendationalgorithmfirstlyadoptsdecisiontreetocalculateaninitialrecommendlist.Itthenusescollaborativefilteringcombinedwithcustomer’purchaserecordtogenerateafinalrecommendationlist.Thefeasibilityoftheproposedrecommendsystemisexperimentallyprovedviaindividualrecommendexperiment.Themethodcansolvethetraditionaldatasparsenessproblemofrecommendationalgorithm.Keywords:RFID;decisiontree;collaborativefiltering;clothingrecommendsystemRFID1,1,11201620Email:[e-mail],[e-mail],[e-mail]tification,RFID1RFID(RadioFrequencyIdentification)RFIDVIPRFIDRFIDNaistenPukutechdasOyNPCollectionRFIDRFIDRFIDRFIDRFID2RFID2.1RFIDRFIDRFID(tag),(reader),(antenna),(application-supportingsystem)RFIDVIP Figure 1. Structure of clothing recommend system based on RFID  1. RFID VIP2.2VIPdoublegender=0gender=11Table1.Quantitativetableofstyle1.   3RFID3.1[3]2Figure 2. Graph of decision tree 2. VIP3 Figure 3. Flow chart of decision tree algorithm 3. 3.2AHP[4]4(O)(C)(1,2,...5)kCk(P)N(1,2,)nPnNFigure 4. Structure of AHP 4. [5]iyjyijaiyjyijaSaaty1-9[6]113/15/17/19/1313/15/17/15313/15/175313/197531A(1)Matlabmax=5.23752(0.5128,0.2615,0.1289,0.0634,0.0333)TW(2)3.3ijijijkind(,)ijcolor(,)ijstyle(,)ijcollar(,)ijprice(,)ij312345kind(,)color(,)style(,)collar(,)price(,)ijijwijwijwijwijw(3)12345(,,,,)44kind(,)kind()kind()ijij(4)3.4[7]2.3ANrecord(1),record(2),,record(n)Mclothes(1),clothes(2),,clothes(m)NM35Figure 5. Flow chart of collaborative filtering algorithm 5. 4200VIP10RFID3333RFID0F0000000000000004VIP412002T75323441236Figure 6. Results of recommended 6. 2Table2.Clothing information recommended2.  ()0.8321(83.21%)0.8012(80.12%)0.7837(78.37%)RFIDRFIDReferences[1]G.Xue,C.LinandQ.Yang,ScalableCollaborativeFilteringUsingCluster-basedSmoothing.InProceedingsoftheACMSIGIRConference,2005.p.114–121.[2]T.GeorgeandS.Merugu,AScalableCollaborativeFilteringFrameworkBasedOnCo-clustering,InProceedingsoftheIEEEICDMConference,2005.[3]W.Hu,XieTongGuoLvSuanFaZaiTuiJianXiTongZhongDeYingYong.ComputerEra,2009,11:11-15(ch)...20091111-15.[4]W.L.LiuandS.Y.Liu,AHPZhongQunJueCePanDuanJuZhenDeGouZao.SystemEngineeringandElectronics,2005,27(11):1907-1908(ch)..AHP.2005,27(11)1907-1908.[5]R.C.TsaiandT.L.Wu,AnalysisofPairedComparisonDataUsingMx.,StructuralEquationModeling,vol.11(2004),p.73-91[6]S.I.GassaandT.Rapcsákin:SingularValueDecompositioninAHP,volume154ofEuropeanJournalofOperationalResearch(2004),p.573-584[7]L.H.UngarandD.P.Foster.ClusteringMethodsforCollaborativeFiltering.InProc.WorkshoponRecommendationSystemsatthe15thNationalConf.onArtificialIntelligence.MenloPark,CA:AAAIPress(1998)