讲座：Data-Driven Asset Selling
题 目：Data-Driven Asset Selling
嘉 宾：Puping Jiang, PhD candidate, Washington University
主持人：花成 助理教授 上海交通大学安泰经济与管理学院
时 间：2021年11月11日（周四） 10:30-12:00
Motivated by online asset selling marketplace business (e.g., used car and real estate), we formulate a data-driven asset selling dynamic pricing framework which utilizes platforms’ access tocustomers’ online behavioral data. Assuming log-concave individual choice function in price and Poisson demand process whose rate may change over a discrete horizon, we show that the model admits ideal properties which facilitate our regret analysis under a dynamic programming setting. To overcome the computational intractability, we propose a deterministic approximation policy (DA policy) and show that DA policy provides an upper bound for the original problem and its induced pricing policy achieves asymptotic optimality as the scale of the problem grows properly. Furthermore, we propose a Thompson-Sampling-based pricing policy (TS) and a Maximum-APosteriori-Estimation-based (MAP) pricing policy when an idiosyncratic latent value in customer utility function can be learned. Since the platform is restricted in an infrequent pricing environment, within each decision epoch, a potentially adequate amount of customer online behavior data can be observed. Utilizing some large-sample deviation properties, we are able to conduct regret analysis on our TS and MAP policies. Finally, we conduct extensive numerical experiments to show that our proposed algorithms can potentially improve the revenue performance significantly compared with an algorithm that is currently implemented by a leading used car platform. Simulations also reveal that active learning (as the TS policy does) may not outperform passive learning policies like MAP policy in our case. This indicates that the effectiveness of exploration highly depends on the nature of the problem, which may be of independent interest.
In the talk, I will also briefly talk about another related on-going work which aims to develop a dynamic selling and acquisition pricing framework for the asset selling platforms from a holistic perspective.
Puping (Phil) Jiang is a PhD candidate in Supply Chain, Operations, and Technology, from Olin Business School, Washington University in St. Louis. His research focuses on revenue management problems in platform settings with applications in asset selling platforms (e.g., used cars, real estate) and retailing platforms, and information issues in risk management (including the contexts of supply chain finance, blockchain implementation, etc.). His paper “Blockchain Adoption for Traceability in Food Supply Chain Networks” is the Winner of 2020 POMS College of PITM Best Student Paper Competition. Before he entered Olin Business School, he got his BS degree in statistics from University of Science and Technology of China (USTC) in 2016. For more information, please visit his website at https://jiang-puping.github.io.