Scholarlink

<p>Trajectory-User Linking (TUL) refers to determining the user to whom a target trajectory belongs and has become an essential trajectory data mining task. Despite significant progress in TUL research using deep learning methods, existing approaches primarily concentrate on basic spatiotemporal features of individual trajectory points. They neglect global spatial correlations, contextual information, and users’ multi-periodic movement patterns, leading to low accuracy in TUL results. A method called Trajectory-User Linking based on Contextual Global Spatial Graph (CGSG-TUL) is proposed. A contextual global spatial graph in CGSG-TUL is constructed based on historical trajectories, which incorporates contextual information such as proximity and category relationships to effectively model spatial correlations of locations. Regarding time encoding, timestamps of check-ins are encoded according to different time scales to capture users’ multi-periodic movement patterns. To obtain the long-term dependencies of the trajectory while reducing noise, a transformer with check-in time relative position encoding and a mask matrix is designed to implement trajectory encoding. Experimental results on Foursquare-NYK dataset and Foursquare-TKY dataset, demonstrate that CGSG-TUL outperforms the state-of-the-art baseline method GNNTUL, with an average improvement of 2.50 and 2.72% in terms of ACC@1 and Macro-F1, respectively.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Trajectory-user linking based on contextual global spatial graph

  • 期刊论文

  • 发表日期：

    2026年4月21日

查看全文
  • Xuan Hou,
  • Lei Zhang,
  • Yutong Wu,
  • Zhi-zhen Liang,
  • Bai-long Liu

摘要

Trajectory-User Linking (TUL) refers to determining the user to whom a target trajectory belongs and has become an essential trajectory data mining task. Despite significant progress in TUL research using deep learning methods, existing approaches primarily concentrate on basic spatiotemporal features of individual trajectory points. They neglect global spatial correlations, contextual information, and users’ multi-periodic movement patterns, leading to low accuracy in TUL results. A method called Trajectory-User Linking based on Contextual Global Spatial Graph (CGSG-TUL) is proposed. A contextual global spatial graph in CGSG-TUL is constructed based on historical trajectories, which incorporates contextual information such as proximity and category relationships to effectively model spatial correlations of locations. Regarding time encoding, timestamps of check-ins are encoded according to different time scales to capture users’ multi-periodic movement patterns. To obtain the long-term dependencies of the trajectory while reducing noise, a transformer with check-in time relative position encoding and a mask matrix is designed to implement trajectory encoding. Experimental results on Foursquare-NYK dataset and Foursquare-TKY dataset, demonstrate that CGSG-TUL outperforms the state-of-the-art baseline method GNNTUL, with an average improvement of 2.50 and 2.72% in terms of ACC@1 and Macro-F1, respectively.