Location Prediction in Social Networks

Abstract

User locations in social networks are needed in many applications which utilize location information to recommend local news and places of interest to users, as well as detect and alert emergencies around users. However, considering individual privacy, only a small portion users share their location on social networks. Thus, to predict the fine-grained locations of user tweets, we present a joint model containing three sub models: content-based model, social relationship based model and behavior habit based model. In the content-based model, we filter out those location-independent tweets and use deep learning algorithm to mine the relationship between semantics and locations. User trajectory similarity measure is used to build a social graph for users, and historical check-ins is used to provide users’ daily activity habits. We conduct experiments using tweets collected from Shanghai during one year. The result shows that our joint model perform well, especially the content-based model. We find that our approach improves accuracy compared to the state-of-the-art location prediction algorithm.