Application of machine learning on plan instability in master production planning of a semiconductor supply chain

Abstract

The progress of digitalization enables new potentials to supply chain management by available data as well as by analysis methods like machine learning. This paper focuses on the master production planning matching demand and supply for a midterm time horizon, in a volatile, diverse and capacity constrained environment. Therefore, a framework for measuring instability is outlined, a machine learning approach to predict instability is developed and applied using the CRISP-DM methodology on real data of a semiconductor manufacturer. The evaluation and results foster the concept and the field of application, but request the next step of prescriptive instability minimization.