Abstract: Among the many individualization recommendation technologies, collaborative filtering with the merit of wide applicability and high quality has been widely used in e-commerce areas. Some scholars also attempted to apply this method in university libraries. However, compared with business system, university library system existed the problems, such as insufficient user information and lack of rating data, which severely impacted the quality of recommendation. In allusion to these problems, a model including both long-term and short-term reading interests of user was created based on user’s characteristics and behavior, which could be used to optimize rating matrix and improve book recommendation quality of collaborative filtering.

Key words: university library