Book recommendation based on community detection

Abstract

In many recommendation systems, the 'best bet' recommendations are shown, but the predicted rating values are not. This is usually referred to as a top-N recommendation task, where the goal of the recommendation system is to find a few specific items which are supposed to be most appealing to the user. While providing academic library services, recommending books for readers is a significant work for constructing personal learning environment. As behaviors in social networks and internet tend to be in groups and the behavior trends are influenced much by the influential entities. In this paper, we firstly propose methods for detecting communities with similar interesting by selecting influential entities. And then propose the recommendation algorithms based on the community detection. At last, by implementation the methods in the real world dataset, our methods perform better than the traditional collaborative algorithms in precision and recall. © 2014 Springer International Publishing.