Community Detection Based on Readers´ Borrowing Records

Academic libraries have recently adopted recommender systems to provide personalized service for increased library-resource use and personalized educations. Readers´ borrowing records help libraries to realize reader preferences and the recommender systems further provide book recommendations for readers. To apply the readers´ borrowing records in recommendations, we give some inceptions in this paper. We find out that most of the people only borrow one or two books once a time and the interval between two successive borrowing records is usually shot as half a month. And by constructing the reader-reader similarity network, we find it have some characteristics: scale-free distribution, the small-world effect and strong community structure. And we propose three different algorithms to detect the communities in the reader-reader similarity network. At last, we compare the proposed algorithms to the existing one on the real world dataset.