How did you get here? Where will you go?

Our technique predicts latent contexts of user movement activities from movement trajectories obtained by using positioning devices equipped in smartphones, etc. This exhibition introduces two methods: transportation mode estimation and destination prediction. (1) Our method estimates transportation modes such as walking, train ,and car by utilizing deep neural networks (DNNs) that automatically extract movement features from trajectory images. This yields better accuracy than existing methods. (2) Our method predicts a user’s destination by modeling human movements using recurrent neural networks (RNNs). This method can achieve better prediction than existing methods because the modeling is robust against data sparsity and can consider long-term dependencies of user’s movement. Our technique enables us to deeply understand user movement activities, which leads to various applications such as personal navigation services and human mobility analysis and control.

Please click the thumbnail image to open the full-size PDF file.