A hybrid forecasting framework with neural network and time-series method for intermittent demand in semiconductor supply chain

Abstract

As the primary prerequisite of capacity planning, inventory control and order management, demand forecast is a critical issue in semiconductor supply chain. A great quantity of stock keeping units (SKUs) with intermittent demand patterns and distinctive lead-times need specific prediction respectively. It is difficult for companies in semiconductor supply chain to manage intricate inventory systems with the changeable nature of intermittent (lumpy) demand. This study aims to propose an integrated forecasting approach with recurrent neural network and parametric method for intermittent demand problems to support flexible decisions in inventory management, as a critical role in intelligent supply chain. An empirical study was conducted with product time series in a semiconductor company in Taiwan to validate the practicality of proposed model. The results suggest that the proposed hybrid model can improve forecast accuracy in demand management of semiconductor supply chain.