Data-Driven Process Optimization for US Supply Chain Resilience Using Machine Learning and SQL

The resilience of the United States supply chain has been critically tested by recent global disruptions, revealing systemic vulnerabilities in forecasting, logistics, and inventory management. This research proposes a robust, data-driven framework that leverages Machine Learning (ML) and Structured Query Language (SQL) to enhance supply chain process optimization and bolster resilience. We developed an integrated data pipeline where SQL was utilized for the efficient extraction, transformation, and loading (ETL) of large-scale, multi-modal data from disparate sources, including ERP systems, IoT sensors, and logistics feeds. Subsequently, ML models, including a Gradient Boosting Regressor for demand forecasting and a Random Forest classifier for risk prediction, were trained on this consolidated dataset. The results demonstrate a significant improvement in forecasting accuracy, with a 23% reduction in Mean Absolute Percentage Error (MAPE) compared to traditional statistical methods. Furthermore, the risk classification model achieved an F1-score of 0.89, enabling proactive identification of potential disruptions in the logistics network. The SQL-driven data infrastructure allowed for real-time querying and monitoring of key resilience indicators, such as inventory turnover and supplier lead time variability. The discussion highlights how this synergistic use of ML for predictive analytics and SQL for scalable data management creates a closed-loop system for continuous process improvement. We conclude that the adoption of such a data-centric approach is imperative for building agile, transparent, and resilient supply chains capable of withstanding future shocks.