|
Activity Number:
|
405
- Winners: Business and Economic Statistics Student Paper Awards
|
|
Type:
|
Topic-Contributed
|
|
Date/Time:
|
Thursday, August 12, 2021 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Business and Economic Statistics Section
|
|
Abstract #317317
|
|
|
Title:
|
Contextual Dynamic Pricing with Unknown Nonparametric Market Noise via Perturbed Linear Bandit
|
|
Author(s):
|
Yiyun Luo* and Will Wei Sun and Yufeng Liu
|
|
Companies:
|
University of North Carolina at Chapel Hill and Purdue University and University of North Carolina at Chapel Hill
|
|
Keywords:
|
Bandit;
Classification;
Contextual Dynamic Pricing;
Unknown Market Noise;
Nonparametric Demand;
Regret Analysis
|
|
Abstract:
|
We consider a contextual dynamic pricing problem in which the seller receives sequential binary purchasing decisions from the customers. A customer's decision depends on both the seller's offered price and the customer's valuation of the product, which involves a random market noise and a linear structure of the product features, marketing environment and customer characteristics. We design a new pricing policy that can tackle both an unknown linear structure and an unknown nonparametric market noise. Specifically, we formulate the linear structure estimation into a classification problem. Given such an estimation, the online pricing problem with unknown market noise and varying contextual information is formulated as a newly introduced perturbed linear bandit. An optimism-based algorithm is proposed for this perturbed linear bandit to balance between market noise learning and greedy pricing. We prove that our policy is able to achieve a T-period regret bound of O(T^(2/3)) against a clairvoyant policy. We also demonstrate the suprior performance of our policy on both synthetic datasets and a real-life Auto-Loan dataset when compared to the state-of-the-art pricing algorithms.
|
Authors who are presenting talks have a * after their name.
