Exploring the Partial Enumeration Problem for Dynamic Hybrid Flowshop Scheduling with Machine Learning

Optimizing resource allocation in dynamic production environments, particularly in Hybrid Flow-Shops (HFS), remains a challenging task. Building on prior work that combined Genetic Algorithms (GA) and Machine Learning (ML) to predict near-optimal allocations (Hamiti et al. (2024)), this study identifies and addresses an issue termed the "partial enumeration problem". This problem arises from impossibility to represent different solutions with equal qualities during data generation, leading to reduced ML models performances. By formalizing the concept, categorizing it, and developing algorithms to detect and mitigate these inconsistencies, by refining training data to account for this issue, this work lays the foundation for future ML models capable of improving decision-making in dynamic and flexible manufacturing systems.