OpenUAS

Open Dataset
Area Embedding
Human Mobility
Urban Computing

OpenUAS is an open dataset that represents urban area usage patterns as vectors, based on large-scale human mobility data from eight major cities in Japan. For each area in a city, it captures characteristics such as office, residential, and commercial districts from patterns of when and how long people stay.

Paper / Dataset / Code / Press Release

Clustering of area embedding vectors in Tokyo (OpenUAS)

Overview

Urban space is often described by administrative boundaries, land use, or facility information, but these alone do not sufficiently capture how people actually use a place.

OpenUAS represents how each area is used as a numerical vector, using stay information extracted from GPS-based mobility data together with Area2Vec. For example, it can analyze areas with long daytime stays on weekdays, areas with concentrated visits on weekends, or areas with many short morning/evening stays — all from human behavior patterns.

OpenUAS also uses an anchoring method so that area embeddings from different cities or different periods can be compared in a shared embedding space. This makes it possible to compare area-usage patterns across cities and over time without sharing raw location data — for example, finding areas similar to a given area in another city, or analyzing temporal changes such as before and after a commercial facility opens.

Dataset

OpenUAS provides the following data:

Area embeddings for 8 major cities in Japan
Trained Area2Vec models
GeoJSON files for map visualization
Cluster labels for 5, 10, and 20 clusters
Cluster visualization HTML built with Kepler.gl
Stay-trend graphs per cluster
Anchor data
Sample data and scripts for training Area2Vec

Clustering of area embedding vectors for each of the eight major cities (OpenUAS)

Dataset name	OpenUAS
Cities	Tokyo, Osaka, Nagoya, Sapporo, Fukuoka, Sendai, Hiroshima, Kyoto
Scale	1.3M+ 50 m meshes
Coverage	~3,300 km²
Input	Stay records extracted from GPS-based human mobility data
Data volume	175M+ stay records
Main outputs	Area embeddings, cluster labels, GeoJSON files, trained models, anchor data
Paper DOI	10.48550/arXiv.2407.19872

Use Cases

OpenUAS is intended for uses such as:

Functional analysis of areas within a city
Cross-city similar-area search
Time-series analysis of urban change
Characterizing commercial, residential, and office areas
Applications to urban planning, store-location strategy, and transportation analysis
Comparing your own location data with open data

Demo / Web App

Coming soon.

Paper

Naoki Tamura, Kazuyuki Shoji, Shin Katayama, Kenta Urano, Takuro Yonezawa, and Nobuo Kawaguchi.
OpenUAS: Embeddings of Cities in Japan with Anchor Data for Cross-city Analysis of Area Usage Patterns.
arXiv:2407.19872, 2024.
DOI: 10.48550/arXiv.2407.19872

@misc{tamura2024openuas,
  title={OpenUAS: Embeddings of Cities in Japan with Anchor Data for Cross-city Analysis of Area Usage Patterns},
  author={Tamura, Naoki and Shoji, Kazuyuki and Katayama, Shin and Urano, Kenta and Yonezawa, Takuro and Kawaguchi, Nobuo},
  year={2024},
  eprint={2407.19872},
  archivePrefix={arXiv},
  primaryClass={cs.LG},
  doi={10.48550/arXiv.2407.19872}
}