需求预测与库存决策的集成研究

Research on the Integration of Demand Prediction and Inventory Decision

大数据背景下的预测方法,为库存决策提供了很多新的解决方案。常见的方法采用先预测、再决策的"两步走"方式,这样每一步考虑的都仅仅是局部最优,不能达到全局最优。考虑将预测与决策结合起来,提出了一个新的FBNV模型,将线性回归和报童问题转变成一个问题,采用线性规划的方式对问题进行求解。同时,使用阿里天池大数据平台上的数据,对模型做了验证,结果证明FBNV能够有效减少库存成本。

With the development of prediction algorithm in＂the era of big data＂,there comes up with many new solutions for inventory decision.While most of the time,researchers treat newsvendor problem as two subproblems,demand prediction problem and inventory descision problem,find the local optimum solutions for each subproblem,and then miss the global optimum.This paper provides a new Feature-Based newsvendor problem（FBNV）model,which integrates demand prediction with inventory descision together and is computable in ＂one-step＂through linear programming method.This paper also apply the algorithms to sales data collected from Alibaba-Tianchi Data platform and the results show that FBNV model can reduce inventory cost efficiently compared with other＂two-step＂methods.