Preprocessing in stochastic programming: The case of capacitated networks

We investigate different preprocessing procedures for stochastic linear programs whose recourse stage involves directed capacitated networks. The procedures are aimed to help the modeling process, but there are important computational implications as well. We discuss how to check if the problem is tight or loose in terms of feasibility, and we demonstrate how to achieve relatively complete recourse. As a result of our developments, we obtain a computationally feasible way of finding all relevant inequalities required by the Gale and Hoffman theorem on feasibility in networks. Computational results are detailed.