Design and implementation of library information circulation service intelligent system oriented to personalized recommendation
Yan Liu · IET conference proceedings. · 2026
Aiming at the shortcomings of traditional library service mode in resource utilization, user demand satisfaction and service efficiency, this paper puts forward an intelligent system of library information circulation service oriented to personalized recommendation. By constructing a multi-dimensional dynamic portrait of users, the system integrates borrowing history, search behavior and social data, and introduces a time decay factor to accurately capture the changes of users' interests. In the recommendation algorithm, a hybrid framework of collaborative filtering (CF), deep learning and knowledge graph (KG) is adopted, which not only mines potential interest groups based on user similarity, but also uses neural network to capture nonlinear relationships, and promotes recommendation diversity through entity relationship reasoning. In addition, the system designed a dynamic feedback mechanism to update the user portrait in real time and adjust the recommendation weight according to the user feedback to optimize the recommendation effect. The experiment is based on the data of a university library in 2023, covering 12,589 teachers and students, 620,000 books and 127 million behavior logs. The results show that the proposed system achieves a click-through rate (CTR) of 23.6% and an average accuracy rate of 0.297 (MAP@10), and the response time is 156ms, In terms of cold start user recommendation, the CTR reached 16.9% on the first day, the retention rate in the first week was 25.7%, and the user satisfaction score was 3.9 (out of 5), The coverage rate of long tail resources reached 42.6%, and its CTR was 9.8%. The system also verifies its importance in maintaining and improving the click-through rate of users by setting different time attenuation factors. Under different concurrency, the system shows good average response time and error rate, which proves its ability of efficient and stable operation. This study provides an effective solution to improve the service efficiency of the library and optimize the user experience, which has important theoretical and practical value.