Machine Learning Risk Scoring for Predicting Supplier Disruptions and Strengthening United States Supply Chain Resilience

Abstract

Global supply chains have become increasingly interconnected, data-intensive, and vulnerable to disruptions arising from financial instability, geopolitical events, transportation constraints, cyber incidents, natural hazards, and operational failures among suppliers. These interconnected dependencies create significant challenges for organizations seeking to anticipate disruptions before they propagate across procurement, production, and distribution networks. Machine learning provides an opportunity to move supply chain risk management from retrospective assessment toward continuous predictive intelligence. This study examines machine-learning risk scoring for predicting supplier disruptions and strengthening supply chain resilience in the United States. It proposes integrating supplier financial indicators, delivery performance, lead-time variability, inventory conditions, geographic exposure, dependency concentration, quality deviations, and external disruption signals into dynamic supplier risk profiles. Supervised learning and ensemble models are evaluated for estimating disruption probabilities and identifying suppliers requiring preventive intervention. The framework further translates predictions into actionable risk tiers supporting supplier diversification, inventory adjustments, alternative sourcing, and contingency planning. Particular attention is given to explainability, data quality, model drift, false-negative disruption predictions, and governance. The study demonstrates how predictive risk scoring can improve early-warning capability while supporting resilient, evidence-driven supplier management.