PhD thesis defense to be held on April 26, 2023, at 12:00 (Old ECE Building, Room 0.2.2)
Picture Credit: Evangelos Theodorou
Thesis title: Intelligent system for optimal inventory management utilizing machine learning algorithms and transfer learning
Abstract: Inventory policy optimization is a critical component of supply chain management. Well planned inventory management can boost operational efficiency and profitability, allowing businesses to meet customer demands while reducing costs. Machine learning algorithms have shown great promise in a variety of domains, including supply chain, in recent years. However, the application of machine learning in the field of inventory policy optimization has been rather limited, with organizations frequently relying on standard simulations to inform their decisions.
To address this gap, this dissertation creates a novel inventory cost minimization framework based on advanced decision-tree based machine learning models. The proposed approach approximates inventory performance at the item level while taking key replenishment policy parameters and demand patterns into account. Several advantages of the proposed approach over traditional inventory simulations include flexibility, adaptability and the ability to quickly compute data-driven approximations. The framework can also incorporate knowledge from items other than those being optimized, which is useful when historical data is limited or heavily influenced by stock-outs.
To evaluate the effectiveness of the proposed framework, the M5 competition dataset was used. The study's findings revealed that the methodology, particularly its transfer learning variant, resulted in significant reductions in total inventory cost while maintaining the same or even improving customer service level. The proposed framework was also incorporated in a decision support system, giving organizations a user friendly and valuable tool aiding in inventory management.
In conclusion, the proposed framework makes an important contribution to the field of supply chain management. The framework provides a highly efficient and effective method for optimizing inventory control settings by leveraging advanced machine learning models. This approach has the potential to transform inventory management practices, resulting in significant improvements in operational efficiency and profitability.
Supervisor: Professor Emeritus Vassilios Assimakopoulos
PhD Student: Evangelos Theodorou