Multi-preference Book Recommendation Method Based on Graph Convolution Neural Network

Abstract

In the book recommendation system, the relationship between users and books can be regarded as a bipartite graph. The user's interest preferences are mined from the graph through Collaborative Filtering recommendation method, and then use the graph convolution neural network to effectively aggregate the characteristics of users and books, so as to form the book recommendation content that user interest. However, the mining of user interest in the existing book recommendation system is always based on a single user preference, ignoring the diversity of user preferences. We propose a multi-preference book recommendation method based on graph convolution neural network to observe the potential reading interest of users when interacting with books. By capturing these reading interests, we can get more information about users’ preferences, so as to recommend books more in line with their preferences. We extract the recommendation dataset from the real scenario of Bohai University Library between May 2014 and May 2021, and evaluate our method on it. The experimental results show that our method effectively improves the performance of book recommendation.