A hybrid multi-criteria decision-making and machine learning approach for explainable supplier selection

Abstract

Supplier selection has become increasingly complex regarding selection criteria caused by expanded data collection processes and supplier numbers due to globalisation effects. This complexity has led to the consideration of Artificial Intelligence (AI) techniques to facilitate and enhance supplier selection. However, the AI techniques most often applied are unfamiliar to stakeholders and have limited explainability, posing a significant barrier to adopting intelligent approaches in supply chains. To address this issue, we propose a hybrid supplier selection framework that combines interpretable data-driven AI techniques with multi-criteria decision-making (MCDM) approaches: the former aims to reduce the complexity of the supplier selection problem, while the latter ensures familiarity to supply chain stakeholders by retaining MCDM at the heart of the supplier selection process. The framework is validated through two real-world case studies supporting supplier selection decisions in oil, gas, and aerospace manufacturing companies. Preliminary results from our case studies suggest that the framework can achieve comparable performance to approaches utilising only machine learning while offering the added benefits of end-to-end explainability and increased familiarity.

Keywords

• Machine learning
• Supplier selection
• Interpretability
• Explainability
• Multi-criteria decision making
• Analytic hierarchy process