Harnessing machine learning for supply chain management: A systematic review and future research framework

초록

This study presents a systematic literature review of 103 Machine Learning (ML) studies in Supply Chain Management (SCM) published between 2019 and 2025, sourced from Scopus, Web of Science, ABI/INFORM, and EBSCO. Descriptive analysis (chronological, geographical, publication source, and ML techniques) was combined with iterative theme identification to examine dominant ML themes and adoption barriers in the Supply Chain (SC) context. ML has emerged as a disruptive technology, offering significant advantages for supply chain planning, execution, and control. Existing reviews have not systematically examined the applicability and adoption barriers of ML in the SC domain, particularly with respect to the integration of Generative Artificial Intelligence (GenAI) and Large Language Models (LLMs). The study identifies critical literature gaps and discusses four major ML themes and 12 sub-themes in SC: (i) demand forecasting, (ii) procurement, (iii) supply chain risk and resilience, and (iv) supply chain network optimisation. The review also identifies technological barriers (model retraining, scalability, and security), psychological barriers (resistance to change and ethics), and contextual barriers (vendor dependency and regulatory compliance). Using the Supply Chain Operations Reference (SCOR) model (Plan, Source, Make, Deliver, Return), this study presents five research propositions and a structured future research agenda, both particularly relevant to the rise of Generative AI and LLMs. It further discusses the technical, social, and managerial implications of ML for supply chain professionals, including comparative performance benchmarks across ML methods and publicly available datasets.

키워드