Data-Driven Two-Stage Distributionally Robust Mean Semi-Variance Mixed-Integer Optimization Model for Location Allocation Problems in an Uncertain Environment

This study considers the uncertainty caused by data asymmetry in supply chains, the risks associated with this uncertainty and the need for robustness in the supply chain network. It discusses the construction of a data-driven two-stage distributionally robust mean semi-variance mixed-integer optimization model to address the location optimization problem under conditions of uncertainty in transportation costs and demand. To solve this model, a distributed separation hybrid genetic algorithm is introduced, enabling determination of the optimal location, distribution strategy and expected return for a distribution center in the worst case. Then, a fresh food supply chain is utilized as a case study to analyze the effects of uncertainty on location allocation decisions while deriving pertinent managerial insights. Additionally, compared to traditional stochastic optimization models, the proposed model demonstrates greater robustness in numerical simulations. The algorithm is also benchmarked against other methods, and its effectiveness and stability are validated in terms of the computational time, the number of iterations and the convergence speed.