Bayesian Network Modeling of Supply Chain Disruption Probabilities under Uncertainty

Abstract

Global supply chains are increasingly susceptible to uncertainties such as natural disasters, geopolitical conflicts, and pandemic outbreaks, resulting in disruptions that incur billions of dollars in annual losses. Traditional methods for modeling disruption probabilities, such as Fault Tree Analysis and Markov Chains, often face challenges in handling multi-source uncertainty, causal ambiguity, and sparse data, limiting their effectiveness in risk prediction. To address these limitations, this study proposes a Bayesian Network (BN)-based framework for modeling supply chain disruption probabilities under uncertainty. First, a multi-dimensional disruption factor system is established, encompassing three key dimensions: external environment (e.g., natural disasters, trade barriers), internal operations (e.g., production failures, inventory shortages), and network structure (e.g., supplier concentration, network density). Second, a hybrid BN structure learning approach is designed, combining expert knowledge elicited through the Delphi method with data-driven algorithms such as the PC algorithm, thereby balancing domain insights with empirical accuracy. Third, BN parameters are learned using maximum likelihood estimation and expert elicitation, effectively addressing data sparsity by integrating historical data with subjective expert judgments. Experimental validation using a real-world dataset from a Chinese automotive component supplier (2018-2023) demonstrates that the proposed BN framework outperforms traditional approaches, achieving a disruption probability prediction accuracy of 89.2%, compared with 76.5% for Fault Tree Analysis and 79.8% for Markov Chains. It also reduces mean absolute error (MAE) by 21.3%-28.7% and provides interpretable causal insights, such as the finding that supplier concentration above 70% increases disruption probability by 42.5%. The framework offers supply chain managers a practical tool to quantify disruption risks, prioritize mitigation strategies, and enhance overall supply chain resilience.