A Deep Learning-Based Fast Route Planning Model for Location-Based Social Networks

Abstract

Nowadays, location-based social networks (LBSN) are a kind of popular media in cyberspace. The LBSN provide geography-aware social services to users, so as to create special social network activities. For users, where to go and how to arrange following positions are the most intuitive problems. As a result, many researchers began to pay attention to the position recommendation in LBSN, in the past decade. However, existing position recommendation methods in LBSN are mostly oriented with situations involving single positions or not many positions. In fact, many scenarios require suggestion of series positions, which is usually ignored by the current research works. To deal with this problem, this paper proposes a deep learning-based fast route planning model for LBSN. Specifically, dependency inside position sequences for each user is modeled with the use of recurrent deep learning model. Then, such deep learning model is used to output prediction results for series future positions for a user. Hence, series positions can constitute the routes for users. Finally, experiments on a real LBSN dataset show the efficiency of the proposed route planning model.