Accurate and Diverse Recommendations via Integrated Communities of Interest and Trustable Neighbors

Abstract

Considering the users' complete spectrum of interests, the limitation of current research on recommender systems lies in that they have only paid attention to improving the accuracy of recommendation algorithms while neglected the diversification of recommendations. In this paper, we integrated a user preference matching algorithm based on communities of interests and a diverse information recommendation algorithm based on trustable neighbors to develop a hybrid information recommendation model that allows for both accuracy and diversity. Results of experiment and evaluation indicated this model can increase the diversity of recommendations with only a minimal accuracy loss.