Extracting Land-Use Patterns using Location Data from Smartphones

Kentaro Nishi (The University of Tokyo), Kota Tsubouchi, Masamichi Shimosaka (The University of Tokyo)

Urb-IoT 2014 - The First International Conference on IoT in Urban Space, 2014/10


Machine Learning Data Science

"This paper proposes an approach to extract area-by-area and daily land-use patterns using location data obtained from users of Yahoo! Japan's smartphone applications. Information used for extracting patterns is extracted from only location data. In this research, a land-use pattern is defined as how the area is used throughout a day. We extract the land-use patterns based on temporal transition in the number of people in the area. In petterns extraction, a clustering technique with an infinite Gaussian mixture model with Dirichlet process mixtures is used, which can be used to discover the appropriate number of patterns. Experiments were conducted in 34 areas over 56 consecutive days. This means 1,904 conditions were studied. The results of our experiments show that our approach successfully extracts land-use patterns using the temporal transition in a population. The results also reveal that additional features which are estimated only from spatio-temporal data helps us control the extracted patterns."