Recommending potential friend in online social networks

Recent years have seen exponential growth of online social networks which provide users with a new communication and information sharing platform. Recommending potential friend to registered users is a critical task that not only helps increase the linkage inside the network and also improves the user experience. However, quantitative approaches for potential friend recommendation in online social networks considering tie strength are missing. In this paper, we therefore propose a novel random walk with restart based approach for potential friend recommendation in online social networks by bringing together network based approach and tie strength, which is composed of two steps. In the first step, a weighted social graph is derived as the basis for potential friend recommendation in online soical networks. In the second step, users' similarity is determined by a novel random walk with restart based similarity measure. We perform an experimental comparison of the proposed method against existing potential friend recommendation algorithms, using a real data set. We show that a significant accuracy improvement can be gained by considering tie strength.