Explainable Artificial Intelligence for Predictive Operational Risk Management: A Hybrid Framework for Manufacturing Systems

• Abiodun Okunola

  Ladoke Akintola University Technology

  Author

The increasing adoption of Artificial Intelligence (AI) in operations management has been constrained by the opacity of advanced predictive models, limiting their practical utility in high-stakes decision environments. This study addresses the critical gap between predictive accuracy and interpretability in operational risk management by developing and validating a hybrid Explainable AI (XAI) framework for manufacturing systems. The proposed methodology integrates XGBoost, Random Forest, and Multi-Layer Perceptron models within a stacked ensemble architecture, augmented with SHAP (SHapley Additive exPlanations) for global and local interpretability. Using a comprehensive manufacturing operations dataset, the framework achieved an accuracy of 89.4% (F1 = 0.88, AUC-ROC = 0.91) in predicting operational inefficiencies, significantly outperforming traditional static Key Performance Indicator (KPI) methods (p < 0.001). The SHAP analysis identified three key predictors—job planning adherence (34%), machine utilization rate (28%), and resource allocation efficiency (22%)—as dominant drivers of operational performance. The framework transforms opaque AI predictions into actionable operational insights, enabling managers to not only anticipate disruptions but understand their root causes. This research contributes a replicable methodology for implementing trustworthy AI in operations, with practical implications for proactive risk mitigation and continuous improvement in manufacturing environments.

• pdf

Copyright (c) 2026 Abiodun Okunola (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.