Supervised learning for integrated forecasting and inventory control
van der Haar, Joost F. Wellens, Arnoud P. Boute, Robert Basten, Rob J.I.
van der Haar, Joost F.
Wellens, Arnoud P.
Boute, Robert
Basten, Rob J.I.
Citations
Altmetric:
Publication Type
Journal article with impact factor
Editor
Supervisor
Publication Year
2024
Journal
European Journal of Operational Research
Book
Publication Volume
319
Publication Issue
2
Publication Begin page
573
Publication End page
586
Publication NUmber of pages
Collections
Abstract
We explore the use of supervised learning with custom loss functions for multi-period inventory control with feature-driven demand. This method directly considers feature information such as promotions and trends to make periodic order decisions, does not require distributional assumptions on demand, and is sample efficient. The application of supervised learning in inventory control has thus far been limited to problems for which the optimal policy structure is known and takes the form of a simple decision rule, such as the newsvendor problem. We present an approximation approach to expand its use to inventory problems where the optimal policy structure is unknown. We test our approach on lost sales, perishable goods, and dual-sourcing inventory problems. It performs on par with state-of-the-art heuristics under stationary demand. It outperforms them for non-stationary perishable goods settings where demand is driven by features, and for non-stationary lost sales and dual-sourcing settings where demand is smooth and feature-driven.
Research Projects
Organizational Units
Journal Issue
Keywords
Inventory, Forecasting, Supervised Learning, Custom Loss Function, Feature Information