Forecasting and optimization of service level in vague and complex SCM by a flexible neural network–fuzzy mathematical programming approach

Abstract

This study presents a flexible meta-modeling ap- proach for modeling and optimization of service level (SL) in vague and complex supply chains. Service level is used as the dependent variable, and ten standard variables including lead time, forecast error, supplier service level, delay, stock cover- age, backlog depth, number of deliverable product, and num- ber of orders are used as independent variables. The proposed approach is composed of artificial neural network (ANN) and fuzzy linear regression (FLR) for optimum forecasting of SL in SCM. Moreover, it compares the efficiencies of FLR, RR, and ANN approaches by mean absolute percentage error (MAPE). The intelligent approach of this study is applied to an actual supply chain system. The case is an international firm, which its responsibility in the supply chain is to distrib- ute electrical and automation products to local outlets. ANN is identified as the preferred model with lowest MAPE and a comprehensive sensitivityanalysis. The proposedapproachof this study is ideal for accurate forecasting of SL in supply chains with possible complexity, ambiguity, and uncertainty. Thiswouldhelpmanagerstoidentifythepreferredpolicywith respect to performance of supply chain in vague and complex environments. This is the first study that presents a flexible approach for accurate prediction of SL in SCM with possible noise, nonlinearity, and uncertainty.