基於灰色理论之短期销售预测方法

[e-mail]-termSaleForecastingbasedonGreyTheoryBeen-ChianChien,Shun-YiLeeShyue-LiangWangDepartmentofInformationEngineeringDepartmentofInformationManagementI-ShouUniversityNationalUniversityofKaohsiungABSTRACTThemainpurposeofbusinessistomakemoney.Formaintainingagoodrelationshipwithcustomersandmaximizingtheproductivityofthecompany,saleforecastingisanimportanttaskinmostoftheERPandSCMsystems.Thetraditionalsaleforecastingusesthetechniquesofpredictionbasedonstatisticaltheory,whichneedlargeamountofhistoricaldata.However,thelifecycleofcommercialproductsareshorterthanbefore.Evenaproducthasdisappearedbeforetheenoughdataforsalearecollectedtosupportforecasting.Thispaperproposesnewmethodsforshort-termsaleforecastingbasedonGreypredictiontheory.WedesigntwoGreypredictionmodelstocompletethetaskofshort-termsaleforecasting.Theexperimentsshowthattheproposedmethodshavebetteraccuracyinshort-termprediction.Besides,thenewmethodsprovideastableandsimplemodelforpredictionincomparisonwiththepreviousmethodsrelatively.Keywords:Customerrelationshipsmanagement(CRM),Greytheory,Greyprediction,Saleforecasting.2(Price)(Quality)(Function)(Delivery)(Service)ERPSAPOracleBaan(1)(2)(3)(3)(1)(2)(Prediction)(Statistics)(Regression)Intel(CPU):(SalesForecastingModel,SFM)3(Subjectiveopinionforecast)(TimeSeries)[2](1)(Juryofexecutiveopinionmethod)(2)(Salesforcecompositemethod)(3)(Users’expectationsmethod)(4)(Indexmethod)()(1)(Movingaverage)(2)(Exponentialsmoothing)(3)(Standarddeviation)(ERP)(1)(Noise)[2][3](Averagingtechniques)(1.1)(1)T1ˆ+tx∑=+=TtttxTx111ˆ(1)4(1.2)(2)∑=+=TtttxTx11)2(ˆ1ˆ(2)(1.3)(Weightedaverage)(3)(1)∑=+=Ttttxx11ˆα(3)zzz(2)[2][3](2.1)(4)(Smoothconstant)(0.1~0.5)n)ˆ(ˆˆ1ttttxxxx−+=+α(4)(2.2)(Trend-adjustedsmoothing)(5))ˆˆ(ˆˆ)1(1)1()1()1()2(−+−+=ttttxxxxα(5)(3)(Standarddeviation)[4](6)5(K)(LT)1ˆ121−−××+=∑=+n)x(xLTKxxnkNkNt(6)(EnterpriseResourcePlanning,ERP)BaanSAP[17][18](SupplyChainManagement,SCM)(1)(2)(3)1510Holt’s1015Winter’s51020Causal102()BoxJenkins50461982[16]IT[6](System)(Relationalanalysis)(Modelconstruction)(Prediction)(Decision)[5]1n))(,),2(),1(()0()0()0(nxxxL=(0)x,GM(1,1)n)(ˆ)0(ξ+nx},2,1{L∈ξξ)(ˆ:)0(ξ+→nxAGOGMIAGO(0)xoo.[6](1)(SequenceGreyprediction)()GM(1,1)(2)(CalamitiesGreyprediction)(3)(SeasonalcalamitiesGreyprediction)(4)(TopologicalGreyprediction)(5)(SystematicGreyprediction)GM(1,1)1(0)x),,3,2,1);(())(,),3(),2(),1(()0()0()0()0()0(nkkxnxxxxLL===(0)x72AGO(1)x(0)xAGO∑∑∑====11211)0()0()0())(,),(),((kknkkxkxkxL(1)x.3)()1(kz)1(5.0)(5.0)()1()1()1(−+=kxkxkz.4:ab∑∑∑∑∑=====−−−−=nknknknknkkzkznkxkznkxkza222)1(2)1(222)0()1()0()1()]([)]([)1()()()1()()(,∑∑∑∑∑∑======−−−=nknknknknknkkzkznkxkzkzkxkzb222)1(2)1(2222)0()1()1()0(2)1()]([)]([)1()()()()()]([.∑==nkkxC2)0()(,∑==nkkzD2)1()(,∑==nkkxkzE2)0()1()()(,∑==nkkzF22)1()]([2)1()1(DFnFnCDa−×−×−−×=,2)1(DFnEDCFb−×−×−×=.5akaeabxekx−−−=+])1()[1()1(ˆ)0()0(.%100)()(ˆ)()()0()0()0(×−=kxkxkxke,)(ke:)()0(kx:)(ˆ)0(kx:8(SaleForecastingModel,SFM)1(1)(Fixedperiodpredictionmodel)(2)(Dynamicperiodpredictionmodel)(1)))(),3(),2(),1(()0()0()0()0(kxxxxL=(0)xkn=+)1(ˆ)0(kxakaeabxe−−−))1()(1()0(ab1m2nn))(),1(,),1(()0()0()0(mxmxnmx−+−=L(0)x,))(,),2(),1(()0()0()0(nxxxL=(0)x.ERPDataPre-ProcessSalesForecastingMarketingStrategyAdjustmentDataWarehouseSFM1:93AGO)1(x)0(x∑∑∑====nkkkkxkxkx1)0(21)0(11)0())(,,)(,)((L(1)x.4)1(5.0)(5.0)()1()1()1(−+=kxkxkz5ab∑∑∑∑∑=====−−−−=nknknknknkkzkznkxkznkxkza222)1(2)1(2)0(2)1()0(2)1(])([)]([)1()()()1()()(,22)1(22)1(2)0(2)1()1(2)0(22)1(])([)]([)1()()()()()]([∑∑∑∑∑∑======−−−=nknknknknknkkzkznkxkzkzkxkzb.6(0)xˆakaeabxekx−−−=+])1()[1()1(ˆ)0()0(.(2)))(),3(),2(),1(()0()0()0()0(kxxxxL=(0)xknn(1)}),,,,{(12654eeeeMinArgnL=.(2))})(,),({(124∑∑=eavgeavgMinArgnL.e%100ˆ×−=(0)(0)(0)xxxe,=+)1(ˆ)0(kxakaeabxe−−−))1()(1()0(,ab101m2en%100)()(ˆ)()0()0()0(×−=nxnxnxen,}),,,,{(12654eeeeMinArgnL=)})(,),({(124∑∑=eavgeavgMinArgnL.3n))(),1(,),1(()0()0()0(mxmxnmx−+−=L(0)x,))(,),2(),1(()0()0()0(nxxxL=(0)x.4AGO(1)x(0)x∑∑∑====nkkkkxkxkx1)0(21)0(11)0())(,,)(,)((L(1)x.5)1(5.0)(5.0)()1()1()1(−+=kxkxkz.6ab∑∑∑∑∑=====−−−−=nknknknknkkzkznkxkznkxkza222)1(2)1(2)0(2)1()0(2)1(])([)]([)1()()()1()()(,22)1(22)1(2)0(2)1()1(2)0(22)1(])([)]([)1()()()()()]([∑∑∑∑∑∑======−−−=nknknknknknkkzkznkxkzkzkxkzb.7(0)xˆakaeabxekx−−−=+])1()[1()1(ˆ)0()0(.2(ne)()(∑neavg)n=4n=52n4567en2.1912.8813.1214.70∑)(neavg11.7311.2312.2112.8911180(PS)72(CS)72SFM(1)38CS2()(α=0.5)7.5%13.3%29.1%27(Movingaverage)(Standarddeviation)(ExponentialSmoothing)CS2CS37867(4,5)()3:(PS1)SampleSizeGreyMovingAverageStandardDeviationExponentialSmoothing49.70%11.14%13.26%12.69%510.46%12.10%13.47%12.74%69.96%12.82%13.60%12.78%710.07%13.18%13.98%12.60%810.70%13.36%14.12%12.39%911.44%13.45%14.24%12.27%1012.07%13.58%14.17%12.19%1112.16%13.76%13.83%12.28%1211.79%14.07%12.99%12.31%Average10.93%13.05%13.74%12.47%122:(PS1)4:(PS2)SampleSizeGreyMovingAverageStandardDeviationExponentialSmoothing416.26%16.57%21.52%18.85%516.36%17.49%21.85%18.78%617.03%18.39%22.11%18.99%717.44%19.02%21.95%18.94%817.39%19.57%21.87%18.95%916.67%20.21%21.08%18.96%1016.79%20.91%20.83%19.07%1116.80%21.78%20.4