Using machine learning for production scheduling problems in the supply chain: A review

• Published in: Computers & industrial engineering Vol. 206; p. 111243
• Main Authors: Ben Hamou, Khalid Ait, Jarir, Zahi, Elfirdoussi, Selwa
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.08.2025
• Subjects: Machine Learning; Optimization; Production scheduling; Supply Chain Management; Production scheduling; Supply Chain Management; Machine Learning; Optimization
• ISSN: 0360-8352
• DOI: 10.1016/j.cie.2025.111243

[Display omitted] •Hybrid approaches with ML improve scheduling flexibility in dynamic environments.•Reinforcement learning dominates SCM scheduling for real-time decision-making.•Supervised and unsupervised learning enhance accuracy and uncover hidden patterns.•Job Shop Scheduling’s predominance reflects its importance in scheduling studies.•Data dependence and computational complexity challenge ML adoption in SCM. Supply Chain Management (SCM) faces significant complexities and challenges in its operational processes, particularly in production scheduling. These challengeshave been the subject of a great deal of research. Machine learning (ML) is widely and successfully used in various fields, including SCM, to help decision-makers cope with complex situations. This article provides a historical overview of research into the application of ML to production scheduling within SCM. It also discusses the major contributions, limitations and future directions of the field. This study shows that (i) the integration of ML algorithms with traditional optimization methods offers significant advantages in terms of flexibility and efficiency for solving complex scheduling problems; (ii) hybrid approaches combining ML techniques with heuristic and metaheuristic methods are particularly effective for dealing with dynamic and uncertain production environments; (iii) although reinforcement learning techniques dominate applications in this field, supervised and unsupervised learning algorithms also play an important role in improving the accuracy and performance of planning models; and (iv) the main limitations identified include dependence on high-quality data, computational complexity, complexity of model generalization, and the difficulty of adapting models to rapid and unforeseen changes in the production environment. Although ML algorithms provide promising solutions for optimizing scheduling processes in SCM, challenges persist, requiring ongoing research to enhance the efficiency, robustness, and interpretability of these systems. Future research should prioritize the development of more efficient hybrid methods, improvements in data quality, and the adaptability of ML models to diverse production environments.