A Learning Model for Inventory of Slow-Moving Items

Probabilistic inventory models assume that demand follows a stable distribution with known parameters. This assumption is reasonable where substantial demand history is available under stable conditions. However, for slow-moving items, such as maintenance items, usually little history is available. In such cases, the assumption of known parameters seems unnecessarily arbitrary. We present here a model that takes into account the uncertainty of the unknown parameters, determines the optimal inventory decision, updates the original distribution assumptions as the passage of time increases our information concerning the parameters, and determines optimal policy.