Personalized Recommendation System for University Digital Libraries Based on Deep Neural Networks

In the era of digital reading, university libraries need to deeply explore library resources and the numerous information generated by interaction with readers in order to meet the differentiated needs of different readers, provide accurate reading recommendation services, and achieve the transformation of libraries from resource capabilities to service capabilities. This will enable more accurate library resource recommendations. This article explores the design of a personalized recommendation system model for university digital libraries based on deep neural networks. By extracting reader and book features, fusing them, and inputting them into a multi-layer neural network to predict the probability of readers borrowing books, the system comprehensively understands users' reading needs and provides accurate and intelligent resource recommendation services for readers. After collecting existing data from a university library, the article constructed a personalized recommendation system and evaluated the recommendation results, indicating that the system can learn and extract user hidden features well, and can to some extent solve the problems of sparsity and cold start in traditional recommendation systems, achieving more accurate, real-time, and personalized recommendations.