An explainable decision model for selecting facility locations in supply chain networks

Suitable facility location selection for customer-required capacity localization is an emerging topic in semiconductor supply chain management. However, this topic has not been thoroughly investigated. For this reason, an explainable artificial intelligence (XAI)-interpreted fuzzy group decision-making (FGDM) approach is proposed in this study to assist a wafer foundry company in selecting suitable facility locations for customer-required capacity localization. The XAI-interpreted FGDM approach aims to overcome the shortcomings of existing visualization tools and techniques for explaining the facility location selection process. To this end, several new visualization tools and methods have been proposed, including hanging gradient bar charts, gradient bidirectional scatterplots, and hanging gradient bar charts for traceable aggregation. After applying the XAI-interpreted FGDM approach to a real case, the new XAI tools enhanced the explainability of the facility location selection process and results. The advantage over the existing XAI tools was up to 36%. In addition, Shapley additive explanations (SHAP) analysis results showed that the factors that impact the assessment results most may be inconsistent with the original judgments of domain experts.