Eksik bilgili ve rassal tedarikli envanter modelleri

The aim of this thesis is to analyze single-item inventory models with random supplyand imperfect information in a random environment. We assume that the randomness insupply is attributed to the random capacity of the producer and/or random availability ofthe transporter. Moreover, the random environment which modulates demand, supply andall cost parameters is modeled as a Markov chain.In the …rst part of the thesis, we assume that the random environment is fully observed.Through the …rst model in this part, we analyze the inventory problem with random sup-ply due to random capacity of the producer and random availability of the transporter.The optimal policy for single, multiple and in…nite planning periods is shown to be anenvironment-dependent base-stock policy. As the second model, we analyze the inventoryproblem with …xed ordering cost and random capacity only. And we present a counter ex-ample showing that environment-dependent (s; S) policy is no longer optimal for this typeof inventory problems.In the second part of the thesis, we assume that the random environment is only partiallyobserved. Therefore, we model the random environment by two processes: an unobservedprocess which is a Markov chain and observed process which is not necessarily a Markovchain. In the …rst model of this part, we analyze same inventory problem as the …rst onein the …rst part; however, the random environment is assumed to be partially observable.Here, we assume that the random capacity as well as random availability is modulated bythe unobserved (real) environmental process whereas all costs are modulated by observedenvironmental process. Then, we present a counter example showing that base-stock policyis not necessarily optimal for this type of inventory problems. By the second model in thispart, we show that base-stock policy is optimal in single and multiple period settings if thecapacity process is observed as costs. However, for this model, analyzing the in…nite periodproblem is not practical because the observed process is not a Markov chain. Therefore, as athird model, we analyze the same inventory problem using su¢ cient statistics. In this case,iwe show that state-dependent base-stock policy is still optimal in single, multiple and in…niteperiod settings. As a fourth model, we analyze inventory problems with unrelaible suppliersand …xed ordering cost by using su¢ cient statistics. We show that state-dependent (s; S)policy is optimal for this type of inventory problems if the availability process of supplieris observable. Finally, we also analyze inventory problems with …nite capacity and randomyield in a partially-observed random environment and show that state-dependent modi…edin‡ated base-stock policy is optimal in single, multiple and in…nite planning periods.ii