Dynamic learning mechanisms in revenue management problems
- One of the major challenges in revenue management problems is the uncertainty of the customer's demand. In many practical problems, the decision maker can only observe realized demand over time, but not the underlying demand. Methods to incorporate demand learning into revenue management problems remain an important topic both for academic researchers and practitioners. In this dissertation, we propose a class of dynamic learning mechanisms for a broad range of revenue management problems. Unlike methods that attempt to deduce customer demand before the selling season, our proposed mechanisms enable a decision maker to learn the demand "on the fly''. That is, they integrate learning and doing into a concurrent procedure and both learning and doing are accomplished during the selling season. Furthermore, we show that our mechanisms are 1) Robust, that is, our mechanisms work well for a wide range of underlying demand structures and do not need to know the structure beforehand; 2) Dynamic, that is, our mechanisms take advantages of increasing sales data as selling season moves forward and update the demand information along the process and 3) Optimal, that is, our mechanisms achieve near-optimal revenues, as compared to a clairvoyant who has complete knowledge of the demand information beforehand. By comparing with different learning mechanisms, we claim that "dynamic learning" is essential for our mechanism to achieve improved performance and the above features. We consider two widely-used models in this dissertation, a customer bidding model and a posted-price model. For each model, we describe a dynamic learning mechanism with theoretical guarantee of near-optimal performance. Although different models entail different mechanism details and different analysis tools, we show that a common denominator is that the improvement of our mechanisms comes from one unique feature: dynamic learning. We also present numerical tests for both models from which the performance of our mechanisms is verified. We believe that the dynamic learning mechanisms presented in this dissertation have important implications for both researchers and practitioners.
|Type of resource
|electronic; electronic resource; remote
|1 online resource.
|Stanford University, Department of Management Science and Engineering
|Glynn, Peter W
|Lai, T. L
|Glynn, Peter W
|Lai, T. L
|Statement of responsibility
|Submitted to the Department of Management Science and Engineering.
|Thesis (Ph.D.)--Stanford University, 2012.
- © 2012 by Zizhuo Wang
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...