ENHANCING SUPPLY CHAIN RESILIENCE ACROSS U.S. REGIONS USING MACHINE LEARNING AND LOGISTICS PERFORMANCE ANALYTICS

Abstract

This study looks at how supply chain resilience can be strengthened across U.S. regions by linking logistics performance analytics with machine learning. We started by framing resilience in measurable terms, focusing on key performance indicators such as on-time delivery, cost efficiency, and variability across carriers and regions. With these metrics in place, the next step was to clean and structure shipment data so that patterns could be revealed. Using that foundation, we built models to predict delays, optimize carrier selection, and detect anomalies that might signal underlying fragility. Forecasting methods were applied to anticipate future shipping costs and route performance, while clustering was used to distinguish between resilient and fragile connections within the network. From there, we moved beyond standard predictive tasks. We experimented with resilience-aware objectives that penalize misclassifying delayed shipments more heavily, tested whether models trained in one region could adapt to another, and subjected the models to stress scenarios that mimicked shocks like surges in demand, noisy data, or carrier disruptions. What stood out is that while traditional metrics and models capture average performance well, resilience-aware methods provide a sharper view of vulnerabilities and recovery capacity. The insights are not just academic. They show that resilience can be operationalized through a combined framework of analytics and machine learning, producing tools that managers and policymakers can use to spot risks earlier, choose more reliable carriers, and plan for disruption. In practice, this makes it possible to see resilience as more than a buzzword: it becomes a measurable, actionable quality of the supply chain that can be managed and improved across U.S. regions.