Dynamic Contextual Pricing with Doubly Non-Parametric Random Utility Models

• DOI: 10.48550/arxiv.2405.06866
• Publication Date: 2024-05-10
• Authors: Elynn Chen; Xi Chen; Lan Gao; Jiayu Li
• Keywords: Parametric model

In the evolving landscape of digital commerce, adaptive dynamic pricing strategies are essential for gaining a competitive edge. This paper introduces novel {\em doubly nonparametric random utility models} that eschew traditional parametric assumptions used in estimating consumer demand's mean function and noise distribution. Existing methods like multi-scale Distributional Nearest Neighbors (DNN TDNN)}, initially designed offline regression, face challenges online due to design limitations, such as indirect observability utility-related variables absence uniform convergence guarantees. We address these with innovative population equations facilitate estimation within decision-making frameworks establish new analytical results on rates DNN TDNN, enhancing their applicability environments. Our theoretical analysis confirms statistical learning distribution minimax optimal. also derive regret bound illustrates critical interaction between model dimensionality smoothness, deepening our understanding under varied market conditions. These contributions offer substantial insights practical tools implementing effective, data-driven strategies, advancing framework models providing robust methodologies navigating complexities modern markets.