Research on Book Recommendation Integrating Book Category Features and User Attribute Information

Aiming at the problem of poor recommendation effect caused by data sparsity in traditional collaborative filtering algorithm in book recommendation, this study proposes a personalized recommendation model (BCF-UAI) that combines book category features and user attribute information, aiming to improve the accuracy and generalization ability of the recommendation system. By introducing user attribute information (gender, age, occupation) and book category features, Embedding technology is used to map discrete attributes into continuous low-dimensional vectors to construct feature representations of users and books. A feature fusion strategy is designed to vectorize and fuse the multi-dimensional attributes of users and the multi-category features of books to generate a unified high-dimensional feature vector. Based on cosine similarity, the potential correlation between users and books is calculated to generate a personalized recommendation list. The experiment uses the Douban book dataset to compare the UCF, ICF and MF, and verifies the effectiveness of the feature module through ablation experiments. The results show that the proposed method is significantly better than the traditional method in RMSE and MAE, and the ablation experiment shows that the contribution of book category features is higher (Precision increased by 3.38%). The research proves that the fusion of user attributes and book categories can effectively mine the deep association of data and alleviate the problem of data sparsity. The research verifies the necessity of attribute information to capture users’ personalized needs, and provides theoretical and practical reference for the optimization of intelligent book recommendation system.