Newsvendor Problem Under The Era Of Big Data

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 sub-problems,demand prediction problem and inventory decision problem,find the local optimum solutions for each sub-problem,and then miss the global optimum.This paper provides a new Feature-Based newsvendor problem(FBNV)model,which integrate demand prediction with inventory decision together and transform the “two-step” policy to “one-step” policy to find the global optimal solution through linear programming method.This paper also applies different algorithms to real sales data collected from Alibaba-Tianchi Data platform and the results show that FBNV model can reduce inventory cost efficiently compared with “two-step” methods or “direct forecasting” method.It also shows that improve prediction accuracy,“two step” methods,“one step” methods can all somehow reduce inventory cost,while “one step” method is the most efficient one that can reduce total original cost by up to 25%.