Twitter, is a popular micro-blogging platform in which users can broadcast brief text updates (of 140 characters or fewer) to the public over the Internet (Bollen, Mao, & Pepe, 2011; Sakaki, Okazaki, & Matsuo, 2010). A status update message, called a tweet, is often used as a message to friends and colleagues (Java, Song, Finin, & Tseng, 2007). One user can “follow” other users, in other words, a user is subscribed to the messages generated by the other users he or she follows. An edge between two users is not necessarily reciprocal: a user can follow another user without being followed back. After its launch on July 2006, Twitter users have increased rapidly: as of 2013, the number of monthly active users is estimated around 241 million worldwide and 500 million tweets are sent per day (Twitter Inc., n. d.). In Mexico, 35% of the population is online, 82% use social media. With over 10.7 million active users and 55 million tweets sent per day (from which 18.3 million come from a mobile device), Twitter has become the 2nd most important social network in Mexico after Facebook (De Choudhury, Monroy-Hernández, & Mark, 2014; El Economista, n. d.; Ramírez, 2012). According to Mexico’s National Institute of Statistics and Geography (INEGI, 2010), 7.879% of all Mexicans live in Mexico City, we estimate the number of Twitter users in Mexico City to be just under one million.
In this work, we present a study of mobility at an individual scale. Our aim is to understand the daily commutes of the Twitter users in the metropolitan region of Mexico City. Our database comprises 4.1 million geo-located tweets obtained from public streams using Twitter’s API. Using clustering algorithms allows us to infer the approximate home-location of 24,135 users and to focus on their day-to-day movements inside the city. By translating this information into a graph and analyzing it, we are able to discern relevant areas of the city. In summary, the contributions we present in this work are: (1) proposing a workflow to translate social-network information into a graph-theoretic framework; (3) present a data-driven separation of a specific urban environment, and (3) study mobility patterns inside Mexico City.
In the following sections we introduce related work on spatiotemporal analysis and human mobility using social networks data, specifically those extracted from Twitter. Then, we present our methodology for data acquisition and mobility graph construction, followed by a detailed analysis of the graph’s properties and future outlook for this work.