加入支持让我们有继续维护的动力！会员畅享查看所有预告
立即购买
管理科学系学术讲座（7月29日）
- 来源：
- 学校官网
- 收录时间：
- 2026-07-14 03:14:49
- 时间：
- 2026-07-29 10:00:00
- 地点：
- 管理学院思源楼524室
- 报告人：
- Dan Zhang（张丹）
- 学校：
- 复旦大学
- 关键词：
- foundation models, revenue management, demand estimation, large language models, data-driven decision-making, operations management, pricing
- 简介：
- Foundation models, large networks pretrained once and applied to new problems via in-context learning, open up new possibilities for revenue management. This talk explores two of them, each drawing on a different foundation model family. In the first, we use a tabular foundation model for demand estimation. Once the set-valued and heterogeneous structure of choice data is encoded, a pretrained model predicts consumer choices with no dataset-specific estimation: on a yogurt scanner panel it improves holdout log-likelihood by 8% and hit rate by 3.6% over hierarchical Bayes while running 16 times faster, with its sharpest advantage at 10 to 40 purchase occasions per consumer, the data-sparse regime typical of scanner panels and revenue management applications. In the second, we use large language models to read a research literature at scale, extracting the decision levers, demand models, and market structures from thousands of papers to map how revenue management has evolved across operations and marketing, from capacity protection to choice-based demand management to algorithmic marketplaces, and to trace where the field is heading. Together, these explorations point to a broad role for foundation models across revenue management research.
- -/- 6
报告介绍：
This talk explores two applications of foundation models in revenue management. The first uses a tabular foundation model for demand estimation, showing significant improvements in accuracy and speed over traditional methods in data-sparse settings. The second applies large language models to analyze thousands of research papers, mapping the evolution of revenue management from capacity protection to algorithmic marketplaces. Together, these studies highlight the transformative potential of foundation models in advancing revenue management research and practice.
报告人介绍：
Dan Zhang is the Associate Dean for Research and Academics and MediaOne Professor of Operations Management at the Leeds School of Business, University of Colorado Boulder. Dr. Zhang’s primary research interests include revenue management, pricing, and data-driven decision-making. He regularly consults with organizations in these areas and is a frequent speaker at academic and industry conferences, corporate events, and academic institutions. He previously served as Chair of the INFORMS Pricing and Revenue Management Section. He has served in editorial leadership roles at several leading journals, including Production and Operations Management, Decision Sciences, Operations Research, Manufacturing & Service Operations Management, and INFORMS Journal on Computing. Dr. Zhang has taught operations management, statistics, and data analytics across undergraduate, graduate, and doctoral programs in in-person, online, and hybrid formats. Since 2016, he has co-developed and taught the Coursera specialization Data Analytics for Business. Dr. Zhang previously served on the faculty of the Desautels Faculty of Management, McGill University, and held visiting appointments at Cornell University, the University of Michigan, and City University of Hong Kong. He was awarded the Changjiang Chaired Professorship by the Chinese Ministry of Education. Dr. Zhang is also a co-founder of Ramsi, where he works at the intersection of artificial intelligence, revenue management, and hospitality, translating academic research into scalable industry solutions.
报告图片：
