Abstract
Understanding human mobility is one of the important but challenging tasks in Location-based Social Networks (LBSN). Recently, a user mobility mining task called Trajectory User Linking (TUL) has become an essential and popular topic, aiming at identifying user identities through exploiting their mobility patterns. Existing methods mainly focus on learning sequential mobility patterns by capturing long-short term dependencies among historical check-ins. However, users have personalized moving preferences, which have not been considered in previous work. Besides, how to leverage the prior knowledge behind human mobility needs to be further investigated. In this work, we present a novel semi-supervised method, called AdattTUL, to make adversarial mobility learning for human trajectory classification, which is an end-to-end framework modeling human moving patterns. AdattTUL integrates multiple human preferences of check-in behaviors and involves an attention mechanism to dynamically capture the complex relationships of user check-ins from trajectory data. In addition, AdattTUL leverages an adversarial network to help in regularizing the data distribution of human trajectories. Extensive experiments conducted on real-world LBSN datasets show that AdattTUL significantly improves the TUL performance.
Highlights
Location-based Social Network apps have widely used in our daily life, e.g., Twitter, Foursquare, Wechat, and etc
We investigate the background of Adversarially Regularized Autoencoders (ARAE) [33] and Wasserstein Generative Adversarial Networks (GANs) [36]
2) COMPUTATIONAL COMPLEXITY There are three main components in our AdattTUL according to Eq (27): (1) Computation between AtEncoder and TDecoder (LAE (φ, ψ)); (2) Computation on Adversarial Net (AdNet) (W(Pφ(T ), Pθ (z))); and (3) Computation on Classification Net (CNet)(LCNet )
Summary
Location-based Social Network apps have widely used in our daily life, e.g., Twitter, Foursquare, Wechat, and etc. As illustrated in Fig., even if there are more than thousands of check-ins in New York from Foursquare dataset and California from Gowalla dataset, the number of different check-ins generated by the users are often small (e.g., less than 100) It reveals that users usually visit a few new checkins, which motivates us to consider the prior knowledge of human visiting preferences before learning their mobility patterns. A recent study indicates that not all check-ins contribute to the dense representation of a given trajectory [16], and human mobility patterns have a highly complex structure. To tackle the concern in trajectory characterization, our AdattTUL uses an end-to-end framework to regularize the latent codes in an adversarial manner It learns underlying information from massive linked and unlinked trajectories to reinforce the model performance and alleviate the data sparsity problem.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
