A study of a personalized library recommendation algorithm

The traditional library recommendation method mainly relies on manual analysis of readers’ reading history, book classification, and user feedback to achieve personalized recommendation, but this method has limitations such as low efficiency and difficulty in accurately capturing users’ complex needs. In view of this, this paper innovatively designs a recommendation system based on Wide & Deep model. The system utilizes a linear model (Wide part) to process hand-constructed features as a way to remember specific user behavioural patterns; at the same time, high-dimensional sparse features are processed with the help of a deep neural network (Deep part) to capture the complex relationship between user interests. After both outputs are further processed by the fully connected layer, the model outputs the probability distribution of each category through Softmax function and then predicts the book categories that users may be interested in. The experimental results show that the model has excellent performance, with an AUC score of over 99%, showing a strong differentiation ability; precision is close to 97%, recall is close to 96%, and F1 value is close to 54%, which not only can accurately recommend the book resources that the user is interested in but also can effectively cover the diversified needs of the user, and provide a better solution for personalized recommendation service in libraries.