Targeted evidence collection for uncertain supplier selection

• Published in: Expert systems with applications 2020-11, Vol.159, p.113583, Article 113583
• Main Authors: Abel, Edward, Cortés Ríos, Julio César, Paton, Norman W., Keane, John A., Fernandes, Alvaro A.A.
• Format: Article
• Language: English
• Subjects: Collection; Component and supplier management; Decision making; Experimentation; Multi-criteria decision analysis; Multi-objective optimization; Multiple criterion; Supplier selection; Suppliers; Supply chain management; Uncertainty
• ISSN: 0957-4174; 1873-6793
• DOI: 10.1016/j.eswa.2020.113583

•An approach to efficient reduction of uncertainty in supplier selection.•Tackling reduction of uncertainly pertaining to suppliers.•Facilitating better informed supplier selection decision making.•Both certain and uncertain criteria can be considered together•Can take into account any number of criteria The problem of selecting which suppliers, and how much of different items to order from each, involves multiple, often conflicting, criteria such as costs and delivery times. Within real world multi-criteria supplier selection problems there is inherent uncertainty involved, and consideration of its impacts and mitigation is a current and important research direction going forward within the field of supplier selection. Uncertainty within multi-criteria supplier selection may be in relation to (i) a decision maker’s ambiguous preferences, such as the importance between criteria, (ii) the suppliers’ supply capacities of, and demand for, different items, and (iii) known information about suppliers with respect to the set of criteria, such as each supplier’s delivery times or their average defect ratios. Whilst previous work has explored the first two of these, less work has explored uncertainty pertaining to information about suppliers in terms of the criteria and, specifically, how it could be efficiently reduced. Such uncertainty is an important problem to address, as it may have a large impact upon an order regarding its perceived quality compared to its realised quality, so reducing such uncertainty can have a significant impact. This paper presents a Targeted Evidence Collection (TEC) approach for efficient reduction of uncertainty, pertaining to suppliers, by looking to efficiently collect additional evidence. The approach looks to utilise and gather evidence intelligently and dynamically – by considering both the likelihood that each supplier will be part of a solution, along with a decision maker’s preferences between criteria – to reduce the uncertainty and efficaciously move towards the most appropriate solution given no uncertainty. The approach is able to handle scenarios for which there are both certain and uncertain criteria present, and can take into account any number of criteria. The TEC strategy is evaluated against alternative approaches, including an active learning based approach, for varying numbers of uncertain criteria, numbers of suppliers, and variations in a decision maker’s preferences. The experimentation highlights how TEC effic