Finite perturbation analysis methods for optimization of periodic (s, S) inventory control systems

We are dealing with single item inventory systems where the period time is constant and the unsatisfied demands are backordered. The demands are independent and identically distributed random variables, but the distribution of those variables are not known. The total cost of a period consists of; ordering cost "K" which is independent of the ordering quantity, holding cost "h" for each item that remains in stock, and penalty cost "p" for the each backordered item. In the considered system, it is known that when the parameters of an (s,S) inventory policy are chosen appropriate, then the expected period cost can be minimized. There are some exact methods or heuristics for finding the optimal s and S parameters in the literature for the case where the demand distribution is known. In our study, we introduce a perturbation analysis based method for finding the optimal s and S parameters where the demand distribution is not known. Our method anticipates the sensitivity of (s,S) parameters to the period cost for the observed demand quantities. This method's performance is compared with a method that uses Integer Programming with the past data and with a method that calculates the mean and standard variation values with the past data and feeds them to the Ehrhardt's Heuristic.