How friends affect user behaviors? An exploration of social relation analysis for recommendation

摘要

Incorporating the influence of social relationships effectively is fundamental to social recommendation (SR). However, most of the SR algorithms are based on the homophily assumption, where they ignored friends’ different influence on users and users’ different willingness to be influenced, which may make improper influence information integrated and harm the recommendation results. To address this, we propose a unified framework to properly incorporate the influence of social relationships into recommendation by the guidance of buddy (friends who have strong influence on user) and susceptibility (the willingness to be influenced) mining. Specifically, the Social Influence Propagation (SIP) method is proposed to identify each user’s buddies and susceptibility and the Social Influence based Recommendation model is proposed to generate the final recommendation. Experiments on the real-world data demonstrate that the proposed framework can better utilize users’ social relationships, resulting in increased recommendation accuracy.