Study on Dynamic Pricing in E-Commerce Using Q-Learning

EasyChair Preprint 15727

Abstract

Dynamic pricing has emerged as a critical strategy in e-commerce, enabling businesses to optimize revenue by adjusting prices dynamically in response to real-time market conditions, customer behavior, and competitor activities. With advancements in machine learning, reinforcement learning (RL) techniques, particularly Q-learning, offer robust tools for developing intelligent dynamic pricing systems. This paper explores the application of Q-learning in dynamic pricing, framing it within the context of a Markov Decision Process (MDP). The pricing agent leverages states, actions, and rewards to iteratively learn optimal pricing strategies that maximize profitability under varying market scenarios. Key advantages of Q-learning include its adaptability to dynamic environments, continuous improvement through data-driven insights, and effectiveness in balancing exploration and exploitation. Empirical studies, such as those by Liu et al. (2021) and Rana and Oliveira (2014), demonstrate significant performance improvements in pricing optimization when Q-learning frameworks are employed. Despite challenges like large state spaces and computational demands, Q-learning remains a promising approach for achieving competitive advantage and scalability in e-commerce dynamic pricing. This review highlights the transformative potential of Q-learning and its ability to handle the complexities of modern market dynamics.

Keyphrases: Adaptive pricing strategies, Customer Segmentation, Deep Reinforcement Learning, E-commerce pricing, Edge resource allocation, Ethical pricing, Non-stationary markets, Q-learning, Reinforcement Learning, machine learning, multi-armed bandit, personalized pricing, pricing optimization, real-time pricing

@booklet{EasyChair:15727,
  author    = {Marappan Sampath and S Vignesh},
  title     = {Study on Dynamic Pricing in E-Commerce Using Q-Learning},
  howpublished = {EasyChair Preprint 15727},
  year      = {EasyChair, 2025}}