A Neural Network Approach to Jointly Modeling Social Networks and Mobile Trajectories

• DOI: 10.1145/3041658
• Publication Date: 2017-08-24T11:49:04Z
• Authors: Cheng Yang; Maosong Sun; Wayne Xin Zhao; Zhiyuan Liu; Edward Y. Chang
• Keywords: Social network (sociolinguistics); Mobile social network; Representation; Social Network Analysis

Two characteristics of location-based services are mobile trajectories and the ability to facilitate social networking. The recording trajectory data contributes valuable resources towards understanding users’ geographical movement behaviors. Social networking is possible when users able quickly connect anyone nearby. A network with location based known as (LBSN). As shown in Cho et al. [2013], locations that frequently visited by socially related persons tend be correlated, which indicates close association between connections behaviors LBSNs. To better analyze mine LBSN data, we need have a comprehensive view each these two aspects, i.e., network. Specifically, present novel neural model can jointly both networks trajectories. Our consists components: construction generation First adopt embedding method for networks: representation derived user. key our lies generating Second, consider four factors influence process trajectories: user visit preference, friends, short-term sequential contexts, long-term contexts. characterize last employ RNN GRU models capture relatedness at short or long term levels. Finally, components tied sharing representations. Experimental results on important applications demonstrate effectiveness model. In particular, improvement over baselines more significant either structure sparse.