Inventory Management using Reinforcement Learning

• Bevan Smith School of Mechanical, Industrial and Aeronautical Engineering, University of the Witwatersrand, Johannesburg, South Africa https://orcid.org/0000-0002-4881-6661
• Arnav Garg School of Mechanical, Industrial and Aeronautical Engineering, University of the Witwatersrand, Johannesburg, South Africa https://orcid.org/0009-0000-3139-4776
• William Rich School of Mechanical, Industrial and Aeronautical Engineering, University of the Witwatersrand, Johannesburg, South Africa https://orcid.org/0000-0002-5013-9815

DOI:

https://doi.org/10.7166/36-1-3133

Abstract

Poor inventory management negatively affects a company’s profits. Too little stock limits potential sales and customer satisfaction, while too much stock increases storage costs and potential damage. This study combined supervised and deep reinforcement learning (DRL) for optimal decision-making that would maximise profits in the supply chain of Company X. The performance was compared with a benchmark heuristic that stocks inventory based on a seven-day forecast. The DRL models achieved a marginally lower net profit, but satisfied significantly lower customer demand compared with the benchmark heuristic, thus showing its potential to help to optimise the supply chain structure, operation, and parameters.