Multi-layered friendship modeling for location-based Mobile Social Networks

Location-based Mobile Social Networks (MSNs) are becoming increasingly popular given the success of Online Social Networks (OSNs), such as Facebook and MySpace, and recent availability of open mobile platforms, such as Apple iPhones and Google Android phones. MSNs extend existing OSNs by allowing a user to know when her friends are around and by providing the ability to meet new people who share her interests. There are few studies, however, on how users are connected through these emerging location-based MSNs. In this paper, we present analysis results of a commercial MSN for which we quantified the correlation between users´ friendship with their mobility characteristics, social graph properties, and user profiles. The evaluation of the derived model from the empirical traces suggests that the model-based friend recommendation is effective, and its performance is better than well-known Naive Bayes classifier and J48 decision tree algorithms. To the best of our knowledge, this paper presents the first study that models the friendship connections over a real-world location-based MSN.