A hybrid path-relinking method for solving two-stage stochastic integer problems

Path-relinking has helped solving deterministic problems by exploring the neighbourhood of elite solutions in an intelligent way. We present an algorithm that combines a mixed-integer linear solver with a truncated pathrelinking method in order to solve two-stage stochastic integer problems with complete recourse and first-stage integer variables. This method takes advantage of a possible scenario-based decomposition in an innovative way. Therefore, path-relinking is used to combine optimised solutions from different scenarios in order to pursue good stochastic solutions. To assess the computational performance of this method we use the stochastic lot sizing and scheduling problem dealing with perishable products. In this problem, firststage decision variables are linked to production sequences and production quantities. After the uncertain demand is unveiled, the second-stage variables decide on the inventory usage. Computational results show a clear advantage of the proposed method when compared to a state-of-the-art mixed-integer linear solver.