EXPRESS: Lead Time Prediction for Inventory Optimization with Machine Learning

Modern decision-support applications build on planning parameters such as lead time, price, yield, etc., which are maintained as master data. The accuracy of master data significantly influences the viability of such applications. However, the maintenance of master data is considered a tedious and error-prone task. In this study, we explore the effectiveness of machine learning techniques to improve the accuracy of plan lead times. We apply both unsupervised and supervised learning methods for creating lead time prediction models. We test our approach using historical data of a global equipment manufacturer. In a numerical analysis the calculated plan lead times are over 30% more accurate than current plan lead times in terms of mean-squared-error (MSE). This increased accuracy of plan lead times reduces inventory investment by approximately 7%.