With the rapid development of intelligent service system in university libraries, personalized academic resource recommendation has become a key technology to improve user experience and resource utilization. This paper improves the traditional two-part relationship graph of user thesis and proposes UAMO model. Combined with Page Rank algorithm and sorting model, the importance of academic resources is quantified. Based on Apriori algorithm, an academic resource aggregation model is established to explore the spatial and temporal correlation between user behavior and resource topics. More than 100,000 resource data and 2,864 user behavior records of a university smart library are selected as experimental data to empirically test the effectiveness of the model. The results of association rule mining show that social science academic resources are browsed by college students in December with the highest confidence level of 21.5933%, and engineering and technology academic resources are browsed by college students in October with a slightly higher confidence level than the minimum confidence level of 10.5357%. The MAP value of this paper’s model (0.3865) is improved by 25.4% compared with the second best model BERT-TextGNN (0.3082), and the MRR value reaches 0.6927, which verifies the feasibility of the model in the intelligent library services of universities and provides technical support for the dynamic recommendation of resources and optimization of subject services.
