Studies On Inventory Management With Partially Observed Supply Capacity In Stochastic Environment

With the rapid development of economic globalization, the companies tend to es-tablish a supply chain network which is cross-regional and even cross-industry. The aimis to obtain a greater economic benefits and market share. Therefore, when the enter-prises begin to change the management model from the traditional extensive economicoperation towards intensive economic operation, the accurate and efcient supply chainmanagement and inventory management, become the key to the business success. In thereal environment, inventory management involves a lot of uncertainty and unobservableinformation, which restricts the development of efective inventory management. In re-cent years, more and more business managers and researchers concerned about the lackof information in the inventory management.In this thesis, inventory management problem with partially observable informationis researched. Based on the situation that there are common cases with partially observedsupply capacity in the steel industry, we study the partially observed supply capacityproblem in the inventory management. The main content of the paper and innovation canbe summarized as follows:This paper first outlines some reasons of the supply capacity nun-observability, andmodels the inventory problem with Markov Decision Processes. Furthermore, weintroduce unnormal probability to simplify the value equation to a linear one. Weanalysis the inventory problem with partially observed supply capacity in the ran-dom environment, and prove the existence of the optimal ordering policies. More-over, compared with the myopic model, we research the impact of partially ob-served supply capacity on the optimal order quantity.Considering in the really inventory management, companies unable to obtain thestate transition probability of supplier’s supply capacity, we formulate this problemusing Bayesian Markov decision process method, based on the observed empiricaldata. The Bayesian estimation method is used to update the distribution function ofthe supply capacity, according to the actual supply amount in each period. Finally,we give the theoretical proof of the existence of the optimal ordering policies, andreach the conclusion that the company will order higher quantity compared in the myopic situation.