Towards Smarter Supply Chains: Machine Learning Methodologies for Delivery Risk Assessment

Supply chain management (SCM) has evolved into a core strategic component for enterprises seeking to gain a competitive edge in an increasingly interconnected global marketplace. With the ever-growing complexity and volume of data available-from procurement and production to distribution and last-mile delivery-organizations are turning to machine learning (ML) and big data analytics to optimize performance and stream-line operations. In this paper, we explore how integrating machine learning into various stages of the supply chain can unlock efficiencies, reduce delivery risks, and enable more informed, data-driven decisions. We employ the DataCo Global Supply Chain dataset as a case study to illustrate how ML classifiers, including K-Nearest Neighbors (KNN), Logistic Regression, Naive Bayes, and Random Forest, can be leveraged to predict late deliveries. Our results show that Random Forest achieved the highest accuracy (over 93%), demonstrating its superior ability to identify at-risk shipments. Furthermore, we emphasize the role of data exploration, feature engineering, and the critical need for addressing data imbalance in ensuring robust model performance. The discussion underscores the transformative potential of ML for improving supply chain efficiency, resilience, and strategic decision-making, while also highlighting future directions to handle evolving customer demands and emerging challenges in an era of rapidly expanding data.

Перевод пока недоступен