Dynamic Pricing Using Thompson Samples in a Multi-Armed Bandit Framework for Increased Market Milking

The present work investigates the use of Thompson Sampling in dynamic pricing techniques, paying special attention to its different expansions to deal with unique problems. It employs four separate approaches — Basic Thompson Sampling, Contextual Thompson Sampling, Non-Stationary Thompson Sampling, and Hierarchical Thompson Sampling — to optimize price choices in a dynamic setting. The use of a multi-armed bandit framework is proposed to improve market milking and, by extension, revenue maximization under a variety of conditions. This research is significant since it is the first to do such a thorough analysis of Thompson Sampling variants for dynamic pricing, providing a nuanced knowledge of their utility in various settings. Contextual Thompson Sampling uses contextual information to help make better pricing selections than the baseline of Basic Thompson Sampling. Both Non-Stationary Thompson Sampling and Hierarchical Thompson Sampling provide for flexible pricing schemes that may respond to changing market circumstances. This adds to the literature by giving a detailed investigation of Thompson Sampling's potential in adaptive pricing, which may be of great use to businesses that are looking to adopt data-driven pricing strategies in today's ever-changing market.