<p>This article is based on the form of “VR + Art Platform” and designs an art management platform that integrates display, trading, and socializing. The development of front-end work is carried out using Unity 3D engine, and modeling software such as 3ds Max is used to model the models used in the system. The backend is developed using IntelliJ IDEA platform, and the SSM framework is built and combined with MySQL database to implement the functions of each module in detail. The system also realizes the interactive connection between front-end scenes and back-end management, and achieves data transmission. When developing for users, platforms need to provide complex recommendation functions. Given that existing recommendation algorithms often perform inaccurately and efficiently when dealing with changes in user preferences and diverse interaction needs on art platforms. This article proposes a research on art management platform recommendation algorithm by combining graph neural network and exhibition design rules. The introduction of attention mechanism enables the graph neural network model to more accurately aggregate important neighbor node information, thereby improving the accuracy of node representation and personalized recommendation effect, and compensating for the shortcomings of traditional recommendation systems in aesthetics and display optimization, making the recommendation results more practical. Compared to existing virtual platforms, the proposed method explicitly highlights its advantages by significantly improving recommendation accuracy through graph neural networks, while simultaneously enhancing display aesthetics and user engagement via integrated exhibition design rules.</p>