Flexible data driven inventory management with interactive multi-objective lot size optimisation

Abstract

We study data-driven decision support and formalise a path from data to decision making. We focus on lot sizing in inventory management with stochastic demand and propose an interactive multi-objective optimisation approach. We forecast demand with a Bayesian model, which is based on sales data. After identifying relevant objectives relying on the demand model, we formulate an optimisation problem to determine lot sizes for multiple future time periods. Our approach combines different interactive multi-objective optimisation methods for finding the best balance among the objectives. For that, a decision maker with substance knowledge directs the solution process with one's preference information to find the most preferred solution with acceptable trade-offs. As a proof of concept, to demonstrate the benefits of the approach, we utilise real-world data from a production company and compare the optimised lot sizes to decisions made without support. With our approach, the decision maker obtained very satisfactory solutions.