Title: Partial Backorder Inventory System: Asymptotic Optimality and Demand Learning

Abstract: We develop a stochastic inventory system which accounts for the limited patience of backlogged customers. While limited patience is a feature that is closer to the nature of unmet demand, our model also unifies the classic backlogging and lost-sales inventory systems which are special cases of the one we propose. We establish the uniform (asymptotic) optimality of the base-stock policy when both demand and patience distributions are known. When the backlogged demands become unobservable, we introduce a novel policy family that operates without backlogged demands information, and prove that it can approach the cost efficiency of the optimal policy in the system when the demand and patience distributions are known. Finally, we consider an online inventory control problem in which backlogged demand is unobservable and demand and patience distributions are also not known, and develop a UCB-type algorithm that yields a near-optimal policy. The regret bounds given by the algorithm are provably tight within the planning horizon, and are comparable to the state-of-the-art results in the literature, even in the face of partial and biased observations and weaker system ergodicity.