MOBILE Communities From Twitter data

Discussion

This map shows a mobility network of commenters on Twitter. To build this network, we collected geocoded tweets (tweets which have a marked location) that were posted between July 16 2020 - August 18 2020. From this collection of tweets, we filtered out tweets of commenters who contributed only once during that time because we wanted to see the movement of various accounts. We also excluded bots and other machine generated tweets (e.g., weather forecasts). From the locations the remaining tweets, we reconstructed a path of each commenter in space. The locations in each path were aggregated with a geohashing algorithm into a grid in which cells have dimensions 152.9*152.4 square meters. The sequences in each commenter’s path were used as connections between cells in the grid of the mobility network. Upon grouping the relationships between cells, we used a Louvain algorithm to detect communities of commenters.

The map shows large and small communities: blue, red, purple and green. The blue mega-community groups communities in the US. The green community can be found in Brazil; the purple - in Japan. All other countries are united into the red community. The map allows the user to choose small communities within each mega-community and inspect their spatial extents, specifically their cores and peripheries. Other, smaller communities have grey-colored markers.

The map clearly shows how countries, states, and provinces formed their own “bubbles” during the pandemic in 2020. In The US, we can identify 20 communities, in Canada - 3, in Brazil -6, and in Japan -7. Particularly interesting are communities that cross boundaries of other countries, as they might contribute to the spread of COVID more than other communities.