Visualization of POI Competitiveness Using Extracted Map Tiles from Social Media Response Since COVID-19

Abstract

Since the spread of COVID-19 around the world, a series of policies and measures are adopted by the Japanese government to control the epidemic. As a result of these policies, people’s daily life and the functional division of society have changed. In order to understand the changes in urban function and people’s daily behavior over the past year, we collected and analyzed over 1.13 million social media data (Twitter in our example) containing geographic information. We propose regional competitiveness, which represents the access frequency of social data in each raster unit to several attributes. In order to analyze the regional competitiveness in different categories and map tiles, we applied an improved spatio-temporal graph attention network model (ST-GAT) based on unstructured POI (point of interest) data and Twitter data in different levels of the map to abstract the city-regional competitiveness. We have developed and evaluated the competitiveness map tiles based on 5 attributes utilized Twitter data at 2020 of Kyoto in Japan. As the spread of COVID-19 disease and government anti-epidemic measures change the frequency of visits to the core of the city and the trend of regional competitiveness, and our results showed that the regional competitiveness in the map tiles obtained by social media data and POI data visualizes the dynamic change analysis of crowd behavior activities and urban social functions. This research enlightens the promising future of spatio-temporal GAT in users’ dynamic responses with geographic information.