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.