多产品多阶段报童问题的在线订购策略

利用集成专家意见的在线序列预测算法研究多产品多阶段报童问题.将任意的一个固定订购量策略看成一个专家意见,用弱集成算法综合考虑众多专家意见构建在线订购策略,并基于弱集成算法的竞争性理论给出在线订购策略的理论保证.首先给出了两产品多阶段报童问题的在线订购策略,证明了其实现的累积收益可与最优固定订购量策略实现的累积收益相当.然后,将两产品多阶段报童问题的在线订购策略及其理论结果推广到多产品多阶段报童问题中.最后在不同的需求类型下,通过数值算例表明构建的在线订购策略相对于最优固定订购量策略具有较强的竞争性能.

On-line ordering strategies for the multi-product multi-period newsvendor problem

The paper uses the online sequential prediction algorithm of aggregating expert advices to study the multi-product multi-period newsvendor problem. Taking any fixed order quantity strategy as an expert advice, this paper utilizes weak aggregating algorithm (WAA) to build online ordering strategies, and provides theoretical guarantees on these strategies based on the competitive theory of WAA. Firstly, this paper provides online ordering strategy for the two-product multi-period newsvendor problem, and proves that the cumulative gain it achieves is as large as that of the best fixed order quantity strategy. Then, the online ordering strategy and its theoretical guarantee for the two-product multi-period newsvendor problem are generalized for the multi-product multi-period newsvendor problem. Lastly, under different demand types, the numerical examples are used to illustrate that the online ordering strategies this paper builds preserve strong competitive performance compared with the best fixed order quantity strategies.