Interpretable Multi-Criteria ABC Analysis Based on Semi-Supervised Clustering and Explainable Artificial Intelligence

Abstract

Multi-criteria ABC classification is an effective technique that allows rapid and automatic organization of a growing number of inventory items into classes having different managerial levels. These built classes help decision-makers efficiently control the inventory and optimize the whole supply chain. However, existing ABC classification methods work as black-box processes that produce ABC classes without providing any explanations behind the assignment of the items. Given the multi-criteria nature of the ABC classification problem, managers cannot easily analyze and interpret the item managerial classes. Another problem of existing methods is their inability to follow the Pareto principle which states that items must be Pareto distributed over the ABC classes. To solve these two problems, we propose a semi-supervised explainable approach based on both semi-supervised clustering and explainable artificial intelligence. The semi-supervised technique is used to integrate an intelligent initialization and a constrained clustering process that guides the classification process to lead to Pareto distributed items, whereas explainable artificial intelligence is used to build detailed micro and macro explanations of the inventory classes at the item and the class levels. Application of the proposed approach for the automatic classification of chemical products of a distribution company has shown the effectiveness of the proposed approach in providing accurate, transparent, and well-explained ABC classes.

Keywords

• ABC classification
• inventory classification
• eXplainble Artificial Intelligence (XAI)
• explainable clustering
• SHapley Additive exPlanations (SHAP)
• constrained clustering