Mining User Behavior and Similarity in Location-Based Social Networks

Location-Based Social Network is a kind of online social network developed on the basis of traditional social network. Location is the cornerstone of its functions and services. The large number of user data that social networks collected provides a more reliable guarantee for exploring and studying the development of human society. Behavior Pattern is some inherent way that can be abstracted and generalized from a large number of actual behaviors. Mining user behavior can find user activities in the law and provide a theoretical basis for many aspects, such as urban planning, commercial distribution, and application development of smart phones. In this work, we mine individual behavior patterns and study user similarity with the real dataset from a typical Location-Based Social Network named Bright kite. We first cluster locations with DBSCAN as a fundamental step. Based on the result of clusters, we mine behavior pattern of each active user exploiting their degree of activity at different location clusters. Further, we propose a method to calculate the user similarity. The analysis of our dataset shows that users of location-based social networks are more willing to do the check-ins at popular locations. We successfully recommend the most similar user for almost all the active users. Besides, the evaluation experiment of user similarity shows that the method we proposed is feasible and effective enough.