Prof. Li Chen
Cornell University
Talk: Data-Driven Contextual Pricing with Semi-Parametric Models
Abstract: We consider a data-driven revenue-maximizing contextual pricing problem with a semi-parametric model of customer valuation (also known as willingness-to-pay). The valuation model is characterized by linear contextual features and a distribution-free noise term. The online learning setting of the problem has been extensively studied in the literature. In this paper, we focus on the relatively understudied offline learning setting of the problem and develop an intuitive estimate-then-optimize algorithm framework. This framework paves the way for one to design asymptotically optimal pricing algorithms and determine their convergence rate with a simple formula based on the underlying statistical estimators of the unknown model parameters, thus bridging our problem to the rich theories and techniques developed in the econometrics, statistics, and machine learning literature. We subsequently apply this general framework to design data-driven pricing algorithms for different data settings and find that estimating an empirical distribution of the unknown noise term has a surprising advantage over many other methods for simplifying the downstream price optimization task. We conduct comprehensive numerical experiments with both synthetic and real data. Tested on synthetic data, our algorithms demonstrate superior performance with various underlying distributions of the noise term. Applied to a real-world online auto loan dataset, our algorithms outperform commonly adopted state-of-the-art benchmarks.
Biography: Li Chen is the Emerson Professor of Manufacturing Management and professor of operations, technology, and information management in the Samuel Curtis Johnson Graduate School of Management, part of the Cornell SC Johnson College of Business. Prof. Chen's research interests include supply chain management, operations strategy, and data-driven analytics. He has published research works in top-tier journals in the operations management field such as Management Science, Operations Research, Manufacturing & Service Operations Management, and Production and Operations Management, where he also serves as a department editor. He is a coauthor of a freely available text analytics R package named STS. Prior to his academic appointments, Chen was a cofounder and lead scientist at TrueDemand Software. He earned his PhD in management science and engineering from Stanford University.
