FORECASTING ON A INTERMITTENT DEMAND

Abstract

Forecasting aamethods aare aaoften aavalued aaby aameans aaof aasimulation aastudies. aaFor aaintermittent aademandaaitems aathereaareaaoftenaaveryaafewaanon–zeroaaobservations, aasoaait aais aahardaato aacheckaanyaassumptions, aabecauseaastatistical aainformationaais aaoftenaatooaaweakaatoaadetermine, aafor aaexample, aadistributionaaof aa aavariable. aaTherefore, aait aaseems aaimportant aatoaaverifyaa theaaforecasting aamethods aaonaatheaabasis aaof aareal aadata. aaTheaamainaaimaaof aatheaarticleaais aanaaempirical aaverification aaof aaseveral aaforecastingaamethods aapplicableaainaacaseaaof aaintermittent aademand. aaSomeaaitems aare aasold aaonly aain aaspecific aasubperiods aa(in aagiven aamonth aain aaeach aayear, aafor aaexample), aabut aamost aaforecasting aamethods aa(such aas aaCroston's aamethod) aagive aanon–zero aaforecasts aafor aall aaperiods. aaFor aaexample, aasummer aawork aaclothes aashould aahave aanon–zero aaforecasts aaonly aafor aasummer aamonths aand aamany aamethods aawill aausually aaprovide aanon–zero aaforecasts aafor aall aamonths aaunder aaconsideration. aaThis aawas aathe aamotivation aafor aaproposing aand aatesting aa aanew aaforecasting aatechnique aawhich aacan aabe aapplicable aato aaseasonal aaitems. aa In aathe aarticle aaeight aamethods aawere aappliedaatoaaconstruct aaseparateaaforecastingaasystems aasuchaas, ▪ CROSTON ▪ TSB_CROSTON ▪ HyperbolicaaExponential aaSmootheningaaModel ▪ SBA ▪ Holt's aaLinear aa ▪ Error aatrendaaSeasonality ▪ L-STARaa(logisticaasmoothaatransitionaaAutoaaregressive) ▪ ARMAaa(AutoaaRegressiveaaMovingaaAverage) The aapresented aanalysis aamight aabe aahelpful aafor aaenterprises aafacing aathe aaproblem aaof aaforecastingaaintermittent aaitems aa(andaaseasonal aaintermittent aaitems aas aawell).