A model for reliable forecasting of supply chain demand with a neural network approach


Safaei M.

Turkish Journal of Computer and Mathematics Education (TURCOMAT), vol.12, no.14, pp.126-131, 2021 (Scopus)

Abstract

Demand forecasting has always been a challenging issue in the supply chain. For this reason, it is  known as the  main  tool  for  success  in  balancing  supply  and  demand.  There  are  many  methods,  such  as  regression,  time series for prediction. If causal relationships between the influential factors of the model are not clear, all of these methods  will  lose  their  accuracy.On  the  other  hand,  considering  all  causal  relationships,  despite  increasing  the accuracy  of  the  model,  makes  it  an  NP-Hard  model.If  the  demand  for  several  customers  is  considered,  solving this  model  will  be  more  difficult  and  sometimes  impossible.In  this  paper,  using  a  combination  of  several artificial  neural  networks  such  as  Principal  Component  Analysis,  Self-Organization  Map,  and  Multi-Layer Perceptron network,  a sustainable hybrid model is  presented.The  purpose of this model is to provide a solutionto  overcome  this  challenge  by  giving  a  reliable  forecast  for  demand,  with  acceptable  accuracy.  The  results  ofthis study all testify to the validity of this claim.

Demand forecasting has always been a challenging issue in the supply chain. For this reason, it is  known as the  main  tool  for  success  in  balancing  supply  and  demand.  There  are  many  methods,  such  as  regression,  time series for prediction. If causal relationships between the influential factors of the model are not clear, all of these methods  will  lose  their  accuracy.On  the  other  hand,  considering  all  causal  relationships,  despite  increasing  the accuracy  of  the  model,  makes  it  an  NP-Hard  model.If  the  demand  for  several  customers  is  considered,  solving this  model  will  be  more  difficult  and  sometimes  impossible.In  this  paper,  using  a  combination  of  several artificial  neural  networks  such  as  Principal  Component  Analysis,  Self-Organization  Map,  and  Multi-Layer Perceptron network,  a sustainable hybrid model is  presented.The  purpose of this model is to provide a solutionto  overcome  this  challenge  by  giving  a  reliable  forecast  for  demand,  with  acceptable  accuracy.  The  results  ofthis study all testify to the validity of this claim.