Who Wants to Join Me?: Companion Recommendation in Location Based Social Networks

Abstract

We consider the problem of identifying possible companions for a user who is planning to visit a given venue. Specifically, we study the task of predicting which of the user's current friends, in a location based social network (LBSN), are most likely to be interested in joining the visit. An important underlying assumption of our model is that friendship relations can be clustered based on the kinds of interests that are shared by the friends. To identify these friendship types, we use a latent topic model, which moreover takes into account the geographic proximity of the user to the location of the proposed venue. To the best of our knowledge, our model is the first that addresses the task of recommending companions for a proposed activity. While a number of existing topic models can be adapted to make such predictions, we experimentally show that such methods are significantly outperformed by our model.