贝叶斯更新下基于情景树的动态订货策略研究

发布时间：2018-07-10 10:09

本文选题：多阶段信息更新 + 情景树　；参考：《南京理工大学》2015年硕士论文

【摘要】：随着世界经济一体化进程的日益加快,网络信息化的飞速发展,商品市场呈现新的特征：客户需求越来越个性化,订货提前期越来越长,产品周期越来越短,需求变得越来越具有可变性。因此,更多的企业将以时间为基础的竞争战略定位为主导竞争态势的重要战略,而快速响应策略(QR)已经逐渐被认可为实现该竞争战略的有效机制。快速响应策略的概念主要是指：通过缩短研究开发、投入生产、销售运营、物流运输等时间来尽量缩短产品的订货提前周期。以便于企业面对客户多变的需求时,能够做出尽可能快的响应,并且根据市场变化调整相应的库存决策。在实际生活中,离销售季节越近,我们能够掌握到的关于不确定性需求的信息越丰富。借此就能使用先进的信息技术(如POS系统等)来收集、分析信息并更新预测需求,最终使得需求不确定性引致的缺货成本或持货成本问题得以缓解。贝叶斯信息更新技术就是比较常用的一种手段,该方法将收集到的市场信息融合进已有的先验信息中,从而修正先验的需求分布,决策者可以根据更新后的需求信息制定一个相对合理的购货计划,将库存量控制在相对适当的水平。在选择缩短运输时间这种方式来延迟订货以获得市场信号进行需求预测更新的情况下,越快的运输模式相应的运输成本将会更贵一些。那么,在运输成本与需求不确定性之间存在一个权衡问题。在本文中,我们基于多重运输模式背景下,探讨涉及需求信息更新的库存问题,研究出一个优化的单一订购策略。从解决最优停止问题的角度,运用动态规划方法来设计最优订购策略。我们以易逝品为研究对象,从零售商的角度考虑如何在恰当的时机选择最优的订货量。本文分别探讨了：在先验需求分布为连续型的情况下,三种运输模式将订货提前期划分为三阶段的动态订货策略；应用情景树理论将预测需求分布离散为模糊近似值的情况下的三阶段动态订货策略；多种运输模式将订货提前期划分为多阶段的动态订货策略。在多阶段模型中,鉴于离散模拟的有效性,我们利用离散化需求分布进行实证演练。我们为每个模型设计相应的动态规划寻优算法,并通过数值仿真和灵敏度分析对相关结论进行论证。
[Abstract]:With the acceleration of the process of world economic integration and the rapid development of network information, the commodity market presents new features: customer demand is becoming more and more personalized, order lead time is getting longer and longer, and product cycle is becoming shorter and shorter. Requirements are becoming more and more volatile. Therefore, more enterprises will take the time-based competitive strategy as an important strategy to dominate the competitive situation, and the rapid response strategy (QR) has been gradually recognized as an effective mechanism to achieve the competitive strategy. The concept of rapid response strategy is to shorten the lead time of order by shortening the time of research and development, production, sales and transportation. In order to meet the changing needs of customers, enterprises can make as fast a response as possible, and adjust the corresponding inventory decisions according to market changes. In real life, the closer we get to the sales season, the more information we can get about uncertain demand. In this way, advanced information technology (such as POS system) can be used to collect, analyze information and update forecast demand, which can alleviate the problem of shortage cost or holding cost caused by uncertainty of demand. Bayesian information updating technique is one of the most commonly used methods, which integrates the collected market information into the prior information, and then modifies the prior demand distribution. According to the updated demand information, the decision-makers can make a relatively reasonable purchase plan and control the inventory at a relatively appropriate level. The faster the transportation mode is, the more expensive it will be if we choose to shorten the transportation time to delay the order to get the market signal and update the demand forecast. So, there is a tradeoff between transportation cost and demand uncertainty. In this paper, we discuss the inventory problem involving the updating of demand information in the context of multiple transport modes, and develop an optimized single ordering strategy. From the point of view of solving the optimal stop problem, the dynamic programming method is used to design the optimal ordering strategy. We take perishable goods as the research object, consider how to choose the optimal order quantity at the right time from the retailer's point of view. This paper discusses the dynamic ordering strategy in which the order lead time is divided into three stages under the condition that the prior demand distribution is continuous. Scenario tree theory is used to discretize the predicted demand distribution into fuzzy approximation for the three-stage dynamic ordering strategy, and various transportation modes divide the lead time into multi-stage dynamic ordering strategies. In the multi-stage model, due to the effectiveness of discrete simulation, we use discrete demand distribution to conduct empirical exercises. We design corresponding dynamic programming optimization algorithms for each model and demonstrate the relevant conclusions by numerical simulation and sensitivity analysis.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：F274

【相似文献】