Implementation of Explainable AI-Driven Framework for Supply Chain Optimization: A Practical Case Study in Industrial Engineering

Abstract

Supply chain optimization remains one of the most complex and data-intensive challenges in contemporary industrial engineering, requiring decision-making frameworks capable of processing high-dimensional operational data while maintaining the transparency and interpretability necessary for practitioner trust and organizational adoption. This empirical study presents the design, implementation, and evaluation of an Explainable Artificial Intelligence (XAI)-driven framework for supply chain performance optimization within a real-world industrial engineering context, addressing the critical gap between advanced predictive model accuracy and actionable operational decision support. A case study methodology was employed within a mid-sized manufacturing organization operating across multiple supply chain tiers, where historical operational data comprising 24 months of procurement, inventory, logistics, and demand records were collected, preprocessed, and used to train a hybrid machine learning model combining XGBoost and Random Forest algorithms for supply chain performance prediction. SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations) techniques were systematically integrated into the predictive framework to generate transparent, human-understandable explanations of model outputs, enabling supply chain managers and industrial engineers to identify the operational variables most significantly influencing inventory inefficiency, delivery delays, and procurement cost overruns. The implemented framework achieved a supply chain disruption prediction accuracy of 94.3 percent, a demand forecasting mean absolute percentage error of 6.7 percent, and an inventory optimization improvement of 28.4 percent compared to the organization's existing rule-based planning system, demonstrating the substantial operational performance benefits of XAI-driven analytical approaches in industrial supply chain environments. Qualitative evaluation through structured interviews with 14 supply chain practitioners confirmed that SHAP-based explanation interfaces significantly enhanced decision-maker confidence, model trust, and willingness to act upon AI-generated recommendations, with 87 percent of interviewed practitioners reporting that explanation transparency was a decisive factor in their acceptance of the framework for operational use. The findings demonstrate that the integration of explainability infrastructure into supply chain ML systems bridges the critical gap between predictive model capability and realized operational improvement, confirming that XAI represents not merely a technical enhancement but a fundamental organizational enabler of AI-driven supply chain excellence in industrial engineering practice.