基于强化学习的智能推荐方法在电子商务中的应用研究

Research on the Application of Intelligent Recommendation Method Based on Reinforcement Learning in E-Commerce

摘要: 推荐系统可以通过推荐用户的个性化项目来减轻信息过载问题。在诸如电子商务之类的现实世界推荐中，系统与其用户之间的典型交互是向用户推荐一页项目并提供反馈；然后系统推荐一页新的项目。为了有效地捕捉这种交互并进行推荐，我们需要解决两个关键问题：(1) 如何根据用户的实时反馈更新推荐策略；(2) 如何生成具有适当显示的项目页面，这对传统的推荐系统提出了巨大的挑战。本文研究了页面推荐问题，旨在同时解决上述两个问题。提出了一种基于深度强化学习的页面推荐框架，该框架能够根据用户的实时反馈，优化页面中的项目，使其在二维页面中得到适当的显示。实验证明了本文方法的有效性。

Abstract: Recommender systems can alleviate the issue of information overload by recommending personalized items to users. In real-world recommendations, such as in e-commerce, the typical interaction between the system and its users involves recommending a page of items to the user and receiving feedback; the system then recommends a new page of items. To effectively capture this interaction and make recommendations, two key issues must be addressed: (1) how to update the recommendation strategy based on the user’s real-time feedback; (2) how to generate item pages with appropriate display, which poses significant challenges to traditional recommender systems. This paper investigates the problem of page recommendation, aiming to address both issues simultaneously. A page recommendation framework based on deep reinforcement learning is proposed. It can optimize the items on the page according to the user’s real-time feedback, ensuring they are appropriately displayed in a two-dimensional page layout. Experimental results demonstrate the effectiveness of the proposed method.