SocialST: Social Liveness and Trust Enhancement Based Social Recommendation
Ran Li, Hong Lin, Yilong Shi, Hongxia Wang · 2019
With the potential value of social relations in recommender systems, social recommendations have attracted increasing attention in recent years. Most existing methods are based on the following assumptions: (1) all users can be affected by their friends easily, and (2) users have similar preferences with those that they trust. However, users who are inactive in social networks may be less influenced by friends, and most people share similar preferences with only a few trusted friends. In this paper, we propose a social recommendation method (SocialST) based on social liveness and trust enhancement. SocialST considers social liveness a key factor in the strength of friends' influence on user preference, and a LivenessRank algorithm is designed to compute social liveness. When modeling social relation, we used a power-growth relationship instead of linear function to simulate the relationship between trust strength and similarity of friend preference, thus strengthening the role of close friends. Experimental results from both the Epinions and Ciao datasets have shown that our algorithm outperforms other current social recommendation algorithms in terms of RMSE.