ICSRec: Interest circle-based recommendation system incorporating social propagation

Collaborative Filtering (CF) is one of the most successful recommendation approaches to overcome information overload. To get a better recommendation, various researches have been conducted in previous literatures. Intuitively, ones' preference may rely on their interest and their friends' suggestion. However, to the best of our knowledge, no existing works systematically combine user's interest preference detection and the influence of social relationship. In this paper, we proposed ICSRec, a novel framework incorporating users' interest groups detection and the influence of social propagation. We first utilize PLSA model to mine the users' and items' interest-circles. In terms of users, a new indicator, POI(point of interest) score, is introduced to measure the extent how a target user is interested in an interest circle. Matrix factorization embedding social propagation is then employed to predict missing preference of a user for an item in each interest-circle. The experimental analysis on two large datasets Epinions and Ciao demonstrates that our approaches outperform other state-of-the-art methods.