商学院青年博士学术工作坊（第二十六期）

讲座题目：Policy Averaging, with an Application to the Newsvendor Problem

主讲嘉宾：Nicholas G. Hall

时间：2025年11月20日（星期四）上午9:30—11:30

地点：商学院304会议室

欢迎感兴趣的师生参加聆听！

江南大学商学院

2025年11月18日

主讲嘉宾简介

Nicholas G. Hall is Berry Family Professor of Operations & Business Analytics, Fisher College of Business, The Ohio State University. He holds B.A., M.A. degrees in economics (University of Cambridge), and a Ph.D. in Management Science (U.C., Berkeley). His research and teaching interests are in project management, scheduling, and sports analytics. He has published over 100 articles in Operations Research, Management Science, Mathematics of Operations Research, Mathematical Programming, Games and Economic Behavior, and others. He has served for over 47 years on the editorial boards of Operations Research and Management Science, and given 456 academic presentations, including 228 invited presentations in 29 countries, 20 conference keynote presentations, and 10 INFORMS tutorials. In 2018, as President of INFORMS, he initiated a government outreach program yielding 30 briefings at The White House and over 500 on Capitol Hill. A 2021 bibliometric study published in Operations Research ranked him 1st in the field of scheduling (1952–2019). He is serving as Editor of the multimedia publication INFORMS Analytics Collections (2021-2026).

讲座主要内容

We propose a Policy Averaging Approach (PAA) that synthesizes the strengths of existing approaches to create more reliable, flexible and justifiable policies for stochastic optimization problems in general. An important component of the PAA is risk diversification to reduce the randomness of policies. A second component emulates model averaging from statistics. A third component involves using cross-validation to diversify and optimize weights among candidate policies. We demonstrate the use of the PAA for the newsvendor problem. For that problem, model-based approaches typically use specific and potentially unreliable assumptions of either independently and identically distributed demand or feature-dependent demand with covariates or autoregressive functions. We show using theoretical analysis, a simulation study, and an empirical study, that the PAA outperforms all those earlier approaches. The demonstrated benefits of the PAA include reduced expected cost, more stable performance, and improved insights to justify recommendations. Beyond the newsvendor problem, the PAA is applicable to a wide variety of decision-making problems under uncertainty.