UNISON data-driven intermittent demand forecast framework to empower supply chain resilience and an empirical study in electronics distribution

Abstract

The complexity involved in demand forecast for supply chain management of electronics components is exponentially increasing owing to demand fluctuations in consumer electronics, shortening of product life cycles, continuous technology migration, lengthy production cycle time, and long lead time for capacity expansion. While global manufacturing networks often suffer the risks of oversupply and shortage of key components, the distributor that is the key intermediate participator in electronics product supply chain buys components from the suppliers, warehouses them, and resells different parts to a number of electronics manufacturers with vendor-managed inventories. Thus, the component distributors forecast the demands for large assortments of stock keeping units (SKUs) with distinct dynamics for inventory control and supply chain management. To address realistic needs to enhance demand forecast performance, this study aims to develop a UNISON data-driven analytics framework that integrates machine learning technologies and temporal aggregation mechanism to forecast the demands of intermittent electronics components. An empirical study is conducted in a world-leading semiconductor distributor for validation. The results have shown practical vitality of the proposed approach with better performance than conventional approaches and the existing practice. Indeed, the developed solution has been employed in this company to support flexible decisions to empower agile logistics and supply chain resilience for smart production.