Science of Machine Learning

Exhibition Program 4

Inferring Pedestrian Flow while protecting privacy

Probabilistic behavior model for discovering pedestrian flow

Abstract

With recent advances in wireless and mobile networks, location information of pedestrians can be recorded in a variety of spaces such as exhibition halls and shopping malls. However, location information of pedestrians is often aggregated for protecting privacy. Aggregated data is a set of incoming and outgoing pedestrian counts at each location. So, it is not straightforward to know pedestrian flow between locations from the aggregated data. In this exhibition, we propose a probabilistic model for inferring latent pedestrian flow between locations using only aggregated data. By incorporating distributions of travel duration between locations, the proposed model can precisely estimate the pedestrian flow between locations. Our model enables us to understand pedestrian mobility patterns while protecting privacy, which provides better navigation and location-based mobile advertising.

Photos

Poster


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

Presenters

Yusuke Tanaka
Yusuke Tanaka
Service Evolution Laboratories
 Hiroyuki Toda
Hiroyuki Toda
Service Evolution Laboratories
Maya Okawa
Maya Okawa
Service Evolution Laboratories
Takuya Nishimura
Takuya Nishimura
Service Evolution Laboratories
Yasunori Akagi
Yasunori Akagi
Service Evolution Laboratories
Shuhei Yamamoto
Shuhei Yamamoto
Service Evolution Laboratories
Takeshi Kurashima
Takeshi Kurashima
Service Evolution Laboratories