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

This study presents a flexible meta-modeling approach 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 coverage, backlog depth, number of deliverable product, and number 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 distribute electrical and automation products to local outlets. ANN is identified as the preferred model with lowest MAPE and a comprehensive sensitivity analysis. The proposed approach of this study is ideal for accurate forecasting of SL in supply chains with possible complexity, ambiguity, and uncertainty. This would help managers to identify the preferred policy with 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.