Viable Supply Chain Network Design: Machine Learning-Derived Chance-Constrained Programming

This dataset contains the computational instances used in the study on two-echelon supply chain network design under facility disruptions. The instances are designed to evaluate models that incorporate cross-echelon dependencies, facility unreliability, and integrated strategies for network viability, including resilience, agility, and environmental considerations. The dataset includes parameterized instances representing different network sizes and configurations, covering small-, medium-, and large-scale problems. Each instance specifies facility locations, customer zones, demand levels, disruption probabilities, capacity limits, fixed and transportation costs, emissions parameters, and service requirements. The data are structured to support the implicit mixed-integer programming formulations in all sizes and also the scenario-based for the small and medium size problems. Additionally, the dataset includes: Generated disruption scenarios for the scenario-based formulation. Training and testing data used for machine learning models that approximate chance constraints. The instances are intended to facilitate reproducibility of the computational experiments and to support further research on reliable and sustainable supply chain network design under uncertainty. Researchers can use this dataset to benchmark optimization models, test decomposition or heuristic approaches, and explore machine learning–enhanced optimization techniques in stochastic and disruption-prone environments.