移动商务环境下服装行业动态销售预测技术研究与实现

上海交通大学硕士学位论文移动商务环境下服装行业动态销售预测技术研究与实现姓名：于天池申请学位级别：硕士专业：机械制造及其自动化指导教师：严隽琪20080101I1.2.3.IIWebServiceABSTRACTIIISTUDYANDIMPLEMENTONDYNAMICSALESFORECASTINGTECHNOLOGIESFORTHEGARMENTINDUSTRYINM-COMMERCEENVIORENMENTABSTRACTThedistributionnetworkofChina’sgarmentindustryalwayshasthreecharacteristics:multi-modes,multi-areasandmulti-levels,whichleadtovariesinformationfromdifferentfieldsexistinginthenetworkandthusit’smoredifficulttomakeadistributionplan.Meanwhile,becauseofclothingproduct’sshortlifecycle,seasonal,regionalandepidemiccharacteristics,theclothingenterprisesmustbuildtheabilityofrapidreactiontomarketchanges.Insuchasituation,howtomakearapidandaccuratesalesforecastingbecomestobethekeyofdistributiondecisioninthesupplychainmanagementofthemoderngarmentindustry.Thearticle’sobjectistoimprovetheaccuracyandrapidityofsalesforecastingofgarmentindustry.Byanalysingthepresentstatus&developmenttrendoftechnologyforclothingsalesforecasting,amethodbasedontheimprovedadjustedmodelisforwarded.CombiningwithrelativetechnologyofM-Commerce,adynamicsalesforecastingmodelisbuilt,relatedsystemisdevelopedanditsapplicationandverificationisimplementedintherealgarmententerprise.Thisarticlemainlyfocusesonthestudyinseveralaspectsasbelow:1.Salesforecastingmethodforthegarmentindustrybasedontheimprovedadjustmentmodel:Firstly,anoptimizedARTmodelisbuilttoimprovetraditionalARmodel,thefeasibilityoferrorpredictionmodelingisalsodemonstrated.Thenmaininfluentialfactorsofthesalesofclothesareintroducedtobuildamultivariableerrorpredictionmodelusingneuralnetwork.Thismethod,whichprovidesadynamicclose-loopforecastingmethodwithfeedbackerror,canimprovetheapplicationofthetraditionaladjustmentmodeltoclothingsalesforecasting.2.SalesforecastingmethodforthegarmentindustrybasedonmobileABSTRACTIVterminals:tobuildadynamicforecastingmodelisputproposed.Thistechnologycollects&exchange&displaytherealtimedistributiondatarapidlyusingmobileterminals.Aimingatthecharictericticsofshort-termclothingsalesforcasting,real-timedataisintroducedtoimprovetheimprovedadjustmentforecastingmodel.Soamulti-stagedynamicforecastingmodelisbuilt,itcomprehensivelyimprovetheaccuracyandrapidityofclothingsalesforecasting.3.ImplementationofthedynamicandaccuracysalesforecastingsystemforthegarmentindustryandVerificationintherealgarmententerprise:Theobjectivedesignandimplementationschemeofthesystemareproposed.Sometechnicaldifficultiesareconquered,suchasWebService,mobilepagedevelopingandforecastingmodelbuildingmethodetc.Theapplicationandverificationisimplementedintherealgarmententerprise.Theresultshowsthatthesystemgreatlyimprovedthequalityofsalesforecastingandsuppliedabetterdecisionsupportforsupplychainmanagementongarmentindustry.KEYWORDS:DynamicSalesForecasting,M-Commerce,AdjustmentModel,ErrorPrediction.20082192008219200821911.11.1.1[1]1-11-1Fig.1-1Typicaldistributionnetworkofclothingenterprises1-121231[2]2312341.1.23DRPDistributionResourcePlanningDRPDRPDRPBSDRPOP3000DRPDRPDRPDRP1DRPDRP2DRPPC1.1.341.21.2.1[3]1ANN[4][5]2GA[4][6]3SVM[5]51[7]2[8]ARMAARMA[9](RBF)RBFRBF[10]AHFCCX390[11-12]16ARMA[13][14][15]DMC[16]ARMA2[17]ARIMA[18]BP[19][20][21][22]FCMAC3[23][24]472080[25]Petra[26]Fischer[27]MingHaiJia[28]1.2.21M-CommercePDA[29]1WLAN2345[30]8GPRS3GIPV6[30]123CPU452[31]SMSWAPIVR[32]Sink[33]RFIDSocketGPS[34]MayTajima[35][36]SyncML[37]PocketPCSQLServer9HTTP[38]J2MEXMLSMSWAP[39][40]SumanKundu[41]GUI1.3123ART102.12-1MRPMPSBOM2-1Fig.2-1Typicalapparelsupplychain2-2111.104.105.206.016.207.057.208.109.109.2010.510.2011.0111.2012.0512.202-2Fig.2-2Atypicalbusinessprocessofgarmentindustry312566040565070122.213−−2-32-3Fig.2-3Frameworkofdynamicsalesforecastingtechniquesforgarmentindustry1142345ERP62.31AutoRegressionTreeARTART23153.13.1.112163.1.21234173.1.33.2Auto-RegressiveandMovingAverageModelARMAARMA12[42]3/18ARTART3-1×××××××××OptimizedParamsErrorErrorARTNeuralNetwork3-1Fig.3-1Structureofimprovedadjustmentmodel3.3ART3.3.1ART19ARTAutoRegressiveAR3-2[43]Milk(0)Bread(0)ARftntnttttXaXaXaXaXε+++++=−−−−L3322113-13-13-2Fig.3-2CasetransformationARTARf3-33-2[43]50006000Milk=3.02+0.72*Bread(t-1)+0.31*Milk(t-1)3-2MonthMilkBreadJan-200550004500Feb-200552004600Mar-200552405130Apr-200563906280May-200567506160Jun-200562806560July-200576807200...CaseIdMilk(t-2)Milk(t-1)Milk(t0)Bread(t-2)Bread(t-1)Bread(t0)15000520052404500456051302520052406390456051306280352406390675051306280612046390675062806280612065605675062807680612065607200...203-3Fig.3-3Regressivetreeoftime-seriesdata3.3.2ART3-3ARTARARTAR[44]1234ARARTART5.2.31COMPLEXITY_PENALTY2MINIMUM_SUPPORT3PERIODICITY_HINT4MISSING_VALUE_SUBSTITUTION5AUTO_DETECT_PERIODICITBread(t-2)=5000Bread(t-2)5000AllMilk(t-1)=6000Milk(t-1)6000Milk=3.02+0.72*Bread(t-1)+0.31*Milk(t-1)21ARTARTART20040120070541200401200605123-13-23-1ART2006/63398795.13529985265412.9409181181929278.75320182006/73023964.3805343455414.56859708072405390.35442792006/83205939.217900855523.79487675462872230.62673702006/93122700.8868368359930.0256474882839724.1597412006/103105953.8905401657283.26494161152626767.98352812006/11304598