The Thirty-Third AAAI Conference on Artificial Intelligence (AAAI-19) will be held January 27 – February 1 at the Hilton Hawaiian Village, Honolulu, Hawaii, USA. The following oral / poster presentation will be delivered.
Analyses of urban dynamics using tons of GPS information from smartphones are has become more common in mobile and pervasive computing.
In this research, we divided the target area into fine-grained meshes and tackled the task to predict active population pattern on each mesh.
Active population prediction on fine-grained meshes helps us understand features of cities more precisely.
However, the smaller number of logs in each mesh and the larger number of meshes for prediction causes the instability in learning.
In this research, we proposed the prediction method which shares the parameters and the datasets among the areas. A spatial preservability of the model is also incorporated by regularization term.
An empirical evaluation with smartphone location data on large-scale fine-grained meshes showed that our model outperforms conventional approaches.
Technical session “Applications and the Web 3” 2019-01-30 (Wed.) 14:00 – 15:30
Spatiality Preservable Factored Poisson Regression for LargeScale Fine-Grained GPS-Based Population Analysis
Masamichi Shimosaka*, Yuta Hayakawa*, Kota Tsubouchi**
(*Tokyo Institute of Technology, **Yahoo! Japan Corporation)
Poster / Demo Session 1 2019-01-29 (Tue.) 18:30 – 20:30