Data-Driven Robust Inventory Management with Time-Series Demand

信息管理与工程学院102室
上海财经大学（第三教学楼西侧）

上海市杨浦区武东路100号

Zhi Chen is an Associate Professor in the CUHK Business School and an Research Associate in CUHK Shenzhen Research Institute, the Chinese University of Hong Kong. His research interests include (1) developing models and designing algorithms for decision-making under uncertainty with different levels of data availability as well as applications in business, economics, finance, and operations; (2) how to compete or cooperate in joint activities such as resource allocation and risk management. His research has been funded by NSFC and Hong Kong Research Grants Council.

We study the multi-period   stochastic inventory management problem with time-series demand in a   data-driven setting. When historical data is limited, the   estimate-then-optimize method often suffers from overfitting and poor   out-of-sample performance. To address this, we propose a data-driven robust   optimization approach that constructs a Wasserstein ambiguity set capturing   demand correlation and uncertainty across the entire planning horizon. We   identify that this modelling approach enables a recursive solution   via robust dynamic programming, and we show that a state-dependent base-stock   policy is robustly optimal. Statistically, we derive finite-sample   performance guarantees for the data-driven robust policy relative to the   full-information optimal policy, extending existing results by explicitly   accounting for demand correlation and distributional uncertainty. Numerical   experiments demonstrate the superior out-of-sample performance of our   data-driven robust policy, particularly with limited data, and underscore the   importance of modeling general time-series demand.