The purpose of this research is to characterize the inventory policies of a two-stage apparel supply chain, consisting of an apparel producer and a retail store, based on the optimal policies obtained by a Markov Decision Process (MDP) and to test the characterized policies on seasonal demand patterns using simulation.
First, the optimal policies for the apparel producer and the retail store were obtained using the MDP experiments. Based on these results, characterized policies for the apparel producer and the retail store were obtained by comparing the MDP results to standard policies such as (Q, r), (s, S), and echelon stock policies. The characterized policies were found to be either optimal or near-optimal. By using the policy characterized for the changes in cost, the total gain of the supply chain was increased compared to using the base policy for that problem. A neural network model found the characterized policies for the apparel producer and the retail store when multiple numbers of costs were changed. The neural network models also obtained near-optimal policies for the models used in this research.
Then, the two-stage apparel supply chain with different types of seasonal demands was simulated. Since the apparel producer and the retail store?s inventory policies were highly correlated to the mean and coefficient of variation (CV) of demand, corresponding policies were required to be changed when the demand had a seasonal pattern in order to increase the total gain of the supply chain. Different options were compared to find the optimal week to change the policy. Option B, which changed policies one week prior to the increase in the mean of demand and changed policies in the same week that the mean of demand decreases, produced the highest gains for both the no information sharing model and the information sharing model. The simulation results showed that by sharing the retail store?s inventory information, both the apparel producer and the retail store could increase their gains. Moreover, the retail store?s gain was increased by a higher percentage than the apparel producer.
