A decision support system for classifying supplier selection criteria using machine learning and random forest approach

Abstract

Supplier selection is an important process in supply chain management that sets a foundation for a long-term partnership with suppliers that can greatly contribute to the success or failure of a business. This study aims to identify, validate and propose a comprehensive list of supplier selection criteria applicable to most organizations. The proposed integrated framework comprises four widely used supervised machine learning (ML) models of Random Forest (RF) classifier and RF-based feature selection algorithm to identify a comprehensive list of critical criteria and their performance measures. We present a case study and show the RF classifier’s performance increased by 3.89% in accuracy and 5.17% in f-score after removing non-critical criteria. Nine criteria are identified as critical among 30 potential criteria considered for supplier selection. Quality, On-Time Delivery, Material Price, and Information sharing are the topmost critical criteria. Another key finding of this study is that transportation cost is a crucial criterion that has received little attention in prior studies. Managers can use this framework to focus on specific criteria when selecting suppliers rather than considering less important criteria or prioritizing the criteria and the suppliers according to their requirements. Many supplier selection studies are reported in the literature, but few studies have utilized machine learning to improve efficacy and effectiveness in supplier evaluation and selection.