Library resource recommendation integrating collaborative filtering and association rules

Abstract

With the explosive growth of digital resources in libraries, the limitations of traditional recommendation systems have become increasingly prominent. The collaborative filtering algorithm is affected by the sparsity of the user-item matrix and is difficult to capture long-tail requirements. The low efficiency of association rule mining leads to serious homogenization of recommendation results. How to break through technical bottlenecks and achieve precise and diverse resource recommendations has become the key to enhancing user experience and promoting the construction of smart libraries. Therefore, a library resource recommendation model that integrates collaborative filtering and association rules is developed to optimize the accuracy and personalization of library resource recommendations. First, the research introduces user attribute features and clustering algorithms to improve collaborative filtering algorithms. Then, a library resource recommendation model is constructed by integrating association rules. Finally, the model is tested. The test results showed that in terms of recommendation diversity, when the number of neighbors was 50, the research model generated 560 unique recommendation items, which was significantly better than traditional content-based recommendation models. In terms of coverage indicators, the research model stabilized at 42%, which was 17% to 30% higher than comparison methods. In real library scenario testing, the research model could identify and recommend six randomly selected user accounts. This research provides an effective technical solution for improving the personalized service level of libraries, which has important practical value for smart libraries.