Relative fragmentation of social activity on the Foursquare social network across New York, London & Paris · Anil Bawa-Cavia (2010)
[ large ]
The literature proposes that the western post-metropolis is characterised by high levels of fragmentation (Soja, 2000 ; Graham & Marvin, 2001). These fragmentation studies show the result of a DBScan clustering analysis of Foursquare activity across New York, London and Paris. You can view a larger version of the graphic. They show how levels of fragmentation detected in the social activity of cities vary greatly across different Western urban settlements.
Foursquare is a location-based social network which users interact with on their 3G mobile phones. The scan-based clustering technique DBScan is used to detect clusters of Foursquare venues within a walkable (400m) distance to each other. Each polygon is formed by joining venues belonging to the same cluster. The darkness of each polygon is proportional to the cluster activity density, calculated as the total number of checkins divided by the number of venues in the cluster. This diagram visualises 845,311 checkins at 18,805 social venues across three cities, showing how urban social activity is splintered into fragments. London and New York show much higher levels of fragmentation, breaking down into almost double the number of clusters than Paris, whose compact form reduces fragmentation, breaking down into far fewer, larger clusters.
Cluster shapes reveal boundaries, for instance forming axial patterns in the case of Paris. These coincide with the larger avenues and boulevards of the city. The Parisian analysis generates under a quarter of the noise of New York, noise in the DBScan context referring to the number of venues that cannot join a cluster due to spatial isolation.
Parisian activity looks far more contiguous using a 400m threshold distance, implying a much more walkable urban fabric.