Geo-Tagged Social Media Data as a Proxy for Urban Mobility

摘要

We evaluate the utility of geo-tagged Twitter data for inferring a network of human mobility in the New York City through a quantitative and qualitative comparison of the Twitter-based mobility network during business hours versus the ground-truth network based on official statistics. The analysis includes a comparison of the structure of the city inferred through community detection in both networks, comparison of the models of human mobility fitted to both networks, as well as the comparison of the dynamic population distribution across the city presented by the networks. Once the utility of the Twitter data is verified, the availability of an additional temporal component in it can be seen as bringing additional value to numerous urban applications. The data visualization web application is constructed to illustrate one of the examples of such applications.