Motivated by the rapid development of information technique, digital resources have become one of the main sources for obtaining information data. Digital library, as the most popular way for obtaining information resources in the network and so on, provides a steady flow of information resources for people on the internet. Further, the authors need to provide personalized recommendation service to users for a recommendation system of digital library research. In the actual system, the sparseness of actual data has a direct influence on the accuracy of the recommendation results. The present methods only improve the accuracy of the recommendation results in a certain extent, but and they cannot be used to solve the problem of caused by data sparseness.