User network construction within online paper recommendation systems

Relationships among users in recommender systems can be presented collectively in the form of user networks. And, the application of user networks is able to provide better personalized service for users. In this paper, we propose a user interest model based on user behaviors, and then we compute user interest similarity based on the interest mode so as to construct user network with similar interest. By processing the metadata of research papers and references, the user network about co-author relationships and the relationships of reference flow can be produced, which we employ to further paper recommendation.