Dynamic Ordering Strategy Research Based On Scene Tree With Bayesian Updates

With the rapid development of network information and the formation of the world economic integration, our commodity market presents new features:customers’ demands become more and more personalized; the order lead time becomes longer and longer; product cycles are getting shorter. They all led to the demand becomes extremely variability. As a result, more and more enterprises treat the competition based on time as the dominant strategy for the market competition. As we know, Quick Response (QR) is an important mechanism to realize the strategy. Quick Response strategy refers to the concept of shortening the order lead time as far as possible by shortening such as research and development time, production time, transportation and sales time. So that, the retailers can make rapid corresponding to customers’demands, and adjust inventory decisions according to the corresponding market change. In real life, the nearer the selling season, we can learn the more information about the uncertain demand, which helps the information technology (such as POS system) to analyze information and update the demand, eventually ease the problem of the shortage cost or the holding cost caused by demand uncertainty. Bayesian infomiation update technology is a common method. It merges the information collected together with the existing prior infomiation to correct the prior demand distribution. Policymakers can make a relatively reasonable purchase plan according to the updated demand information to control the inventory in a relative appropriate level.If we choose the way of delaying order by shortening the transport time to obtain the market signal for demand forecast updating, the faster transportation mode means the corresponding transportation cost will be more expensive. Then, there exists a balance problem between the transportation cost and the demand uncertainty. In this article, we study the inventory problem involving the demand infomiation updating and design a single order optimization strategy based on multiple transportation modes. We use dynamic programming method to design the optimal order strategy from the perspective of solving the problem of optimal stopping. We choose the perishable goods as the research object, consider the right time to choose the optimal order quantity from the perspective of retailer. This paper respectively discusses the following three models:the dynamic ordering strategy a continuous priori demand distribution, three kinds of transport mode dividing the order lead time into three stages; the three stages dynamic ordering strategy in the case of predicting the demand distribution by the discrete fuzzy approximation from scenario tree theory; the dynamic ordering strategy in the case of various transportation modes dividing the order lead time into multiple stages. In the multi-stage model, for the effectiveness of the given discrete simulation, we use the discretization demand distribution to conduct the empirical exercise. We design the corresponding dynamic programming optimization algorithm for each model, and demonstrate the related conclusion through the numerical simulation and sensitivity analysis.