demand forecasting for inventory control filtering outliers

《demand forecasting for inventory control filtering outliers》由会员上传分享，免费在线阅读，更多相关内容在工程资料-天天文库。

1、Chapter9FilteringOutliers9.1IntroductionAprimarygoalofforecastingistomeasuretheflowofdemandsfromthehistorymonthsandprojecttothefuturemonthswithaminimumforecasterror.Awaytoenhancethisgoalisbyfilteringthehistorydemandstoseekoutanyoutlierdemandsandadjustacco

2、rdingly.Asdemonstratedinthepriorchapter,outlierdemandscausemuchdamagetotheforecastsandincreasetheforecasterror.Filteringofthede-mandhistoryisnotaneasyprocess,butisimportanttoyieldforecastswithminimalforecasterror.Reducingtheforecasterrorwillreducetheamoun

3、tofsafetystockneededtoruntheinventoryoperation.Thischaptershowsawaytoseekoutandadjustanyoutlierdemandsfromthehistorymonthswhenthedemandpatternsareofthehorizontal,trendorseasonaltype.Thefilteringprocesstakesplacejustpriortogeneratingtheforecasts.Anotherway

4、tominimizeoutlierdemandsisbyfilteringthelinedemandsthatoccurintheorderentryphaseoftheinventorystockinglocation.Thisiswhenthecustomerssendinthepurchaseordersandlistoneachlineoftheorder,apartnum-berandaquantity.Thispartnumberrequestisalinedemandandcanbefilt

5、eredtoseekifitisanoutlier.Ifanoutlierisdetectedhere,thelinedemandissentbacktothecustomerseekingverificationonthequantity.9.2HorizontalFilteringSupposethehorizontaldemandpatternwhereNhistorydemandentries,x(1),…,x(N),aretobeusedtogenerateahorizontalforecast

6、.Priortoforecasting,thede-mandsarefilteredseekingifanyoutlierdemands,andifso,theoutlierisadjustedaccordingly.Awaytodothisisdescribedinthefollowingsevensteps.©SpringerInternationalPublishingSwitzerland2015119N.T.Thomopoulos,DemandForecastingforInventoryCon

7、trol,DOI10.1007/978-3-319-11976-2_91209FilteringOutliers9.2.1HorizontalFilteringAlgorithm(HFA)1.Foreachx(t),findtheabsoluteaveragedifference,calledthedifferencemeasureanddenotedasd(t),fromitsclosestneighborsasfollows:dt()=

8、

9、xt0.5xt1xt1()−(−++)()fort2t

10、oN1=−d(1)

11、x(1)x(2)

12、fort=−=1d(N)=−−

13、x(N)x(N1)

14、fort=N2.Locatethemaximumd(t),denotedasdmx,andcallthemonthtmxwheretheassociateddemandisx(tmx).3.Formonthtmx,gettheadjusteddemandastheaverageofitsclosestneighborsasbelow:x`