Deep Multiple Instance Learning for Human Trajectory Identification
Zipei Fan*, Xuan Song*, Quanjun Chen*, Renhe Jiang*, Kota Tsubouchi and Ryosuke Shibasaki* (* The University of Tokyo)
27th International Conference on Advances in Geographic Information Systems (ACM SIGSPATIAL 2019), 2019/11
機械学習 (Machine Learning) データサイエンス (Data Science) その他の取り組み (Misc.)
- Extracting identifiable information from human trajectories is a fundamental task in many location-based services (LBS), such as linking heterogeneous sources of data, personalized POI recommen- dation system, irregular human movement detection and privacy protection. However, the multimodal mobility patterns underlain in human trajectories are difficult to model by many existing mod- els. Moreover, we could hardly define a clear user set for user identification because the set of users are dynamic and changing everyday. Bearing these in mind, we apply a deep multiple instance learning method to handle the multimodal mobility patterns in a weak-supervised learning way, and address the dynamic user set problems via a pairwise loss with negative sampling. We utilize a multi-head attention mechanism to automatically extract multiple aspects and match the corresponding information between query trajectories and historical trajectories (document trajectories). Our method shows a good potential on a human trajectory modeling applications such as irregular trajectory detection and sensitivity estimation, and also shows a good identification accuracy on three human GPS trajectory data sets comparing with baseline methods.