Distribution-Free Inventory Risk Pooling in a Multilocation Newsvendor

We study a multilocation newsvendor network when the only information available on joint distribution of demands are values its mean vector and covariance matrix. adopt distributionally robust model to find inventory levels that minimize worst-case expected cost among distributions consistent with this information. This problem is NP-hard. closed-form tight bound there two locations. under family demand six support points. For general case, we develop computationally tractable upper if costs fulfilling have nested structure. optimal value semidefinite program whose dimensions polynomial in number propose an algorithm can approximate fulfillment structures by structures, yielding heuristic for optimization networks. conduct experiments networks resembling U.S. e-commerce show approach over stochastic assumes incorrect distribution. paper was accepted Chung Piaw Teo, optimization.