A Predictive Approach to Define the Best Forecasting Method for Spare Parts: A Case Study in Business Aircrafts’ Industry

Abstract

The cost-effective management of spare parts is an important objective for all manufacturing and service companies. One of the most difficult challenges, for this objective, is accurate demand forecasting and optimized supply planning decisions to achieve best availability level for the spare parts. The main objective of this paper is to propose a predictive approach to identify the best forecasting method with least error cost. Moreover, in business aircraft industry the best forecasting method for a part can change due to the high-level uncertainty in demand. To this purpose, a methodology to select the best forecasting method based on binary classifier machine learning is developed. Proposed methodology is applied in a real case for a well-known business aircraft. The results indicate that neural network is the best machine learning method for 98% of demand and random forest is the best machine learning method for only 2% of parts.