Artificial Intelligence in Supply Chain Optimization - A Systematic Review of Machine Learning Models, Methods, and Applications - Abstract

Aug 5, 2026

Modern supply chains face immense complexity and uncertainty, prompting the integration of Artificial Intelligence and Machine Learning for data-driven prediction and mathematical optimization to support robust decisions. A systematic literature review analyzes 199 articles on tangible supply chains, examining how ML is used—most applications either predict model parameters or directly generate solutions. The study proposes a taxonomy covering ML roles, problem types, and optimization models, synthesizes trends including reinforcement learning in logistics formulations, and identifies research gaps with a new guiding framework for future work.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages68

Reading Time2 h 51 min

AccessFree to download

Abstract

1 Introduction

Artificial Intelligence in Supply Chain Optimization: A Systematic Review of Machine Learning Models, Methods, and Applications December 1, 2025 Nipun GalanI.deE1s,raKBim¨uyiy¨uaktMoahthaamkınm1*adi Jozani 1 , 1 Grado Department of Industrial and Systems Engineering, Virginia Tech, Blacksburg, VA, 24061, USA. *Corresponding author(s). E-mail(s): [esratoy@vt.edu](esratoy@vt.edu) ; Abstract Modern supply chains face immense complexity and uncertainty, driving the powerful integration of Artificial Intelligence (AI) and Machine Learning (ML) for data-driven prediction and mathematical optimization to enable robust decisionmaking. To understand the state of this critical intersection, this paper presentsa systematic literature review analyzing 199 articles focused on Tangible Supply Chains. Our analysis also categorizes how ML is used, with most applications falling into either predicting model parameters or directly generating solutions. This paper introduces a comprehensive taxonomy spanning ML roles, problem types, and optimization models, and synthesizes trends, such as the dominance of Reinforcement Learning in logistics-focused formulations. Beyond providing a detailed classification of the field, this review highlights critical research gaps and contributes a novel research framework designed to guide researchers. This work serves as an essential resource for understanding current trends and identifying future opportunities at the confluence of ML and supply chain optimization. Keywords: Supply chain, Optimization, Machine learning, Logistics 1 1 Introduction The design and optimization of supply chains are fundamental pillars of the modern global economy, directly influencing business competitiveness, sustainability, and social resilience. In an increasingly complex and volatile world, the ability to make optimal decisions regarding production, inventory, and logistics is no longer a competitive advantage but a necessity for survival [62] . For decades, the field of Operations Research has provided the foundational tools for this endeavor through mathematical optimization, enabling organizations to solve complex planning and scheduling problems. The challenges in this domain are multi-faceted, involving multi-scale, multiobjective [1], and multi-player considerations. Addressing these challenges through rigorous optimization is vital for achieving operational excellence and navigating the intricate trade-offs between cost, service, and sustainability. However, having the right tool does not guarantee the successful optimization of supply chain operations. It is important to know how to use the ML tool according to the best practices. Through the strategic implementation of ML algorithms in supply chain optimization, a supplier’s competitive advantage can be determined. In parallel to mathematical optimization, the recent boom in data has led to the introduction of a new direction for solving supply chain problems. The modern supply chain produces streams of information from sensors, transactions, and external sources. To make the best use of these data, organizations are turning to ML models for every aspect, ranging from prediction to improving solutions. ML algorithms can identify the underlying patterns and break down complex data for prediction, such as the prediction of demand [175], without being explicitly programmed. Similarly, ML is also being used to solve large-scale supply chain problems [2] that take days to solve using traditional methods. In more advanced applications, ML models can even help in the formulation of the optimization problems themselves by learning complex, data-driven constraints, or objective functions. The rapid involvement of ML models suggests a paradigm shift, moving supply chain problems from traditional models to incorporating learning methods. While optimization provides the framework for decision-making and machine learning offers data-driven insights, the tr

FAQs

What is the “predict-then-optimize” paradigm mentioned in the introduction?

It is an integration approach where ML provides real-time predictions of uncertain variables, and those predictions feed into optimization models instead of relying on static historical averages.

Aug 5, 2026

Upload by Valentina