Urban Data Mining : Social Media Data Analysis as a Complementary Tool for Urban Design

The emergence of "big data" has resulted in a large amount of information documenting daily events, perceptions, thoughts, and emotions of citizens, all annotated with the location and time that they were recorded. This data presents an unprecedented opportunity to help identify and solve urban problems. This thesis aimed to explore the potential of machine learning and data mining in finding patterns in "big" urban data. We explored several different types of user generated urban data, including Call Detail Records (CDR) data and social media (Crunch Base, Yelp, Twitter, and Flickr, and Trip Advisor) data on two primary urban issues. First, we aimed to explore an important 21st century urban problem: how to make successful "Innovative district". Using data mining, we discovered several important characteristics of "innovative districts". Second, we aimed to see if big data is able to help diagnose and alleviate existing problems in cities. For this, we focused on the city of Andorra, and discovered potential reasons for recent declines in tourism in the city. We also discovered that we can learn the travel patterns of tourists to Andorra from their past behavior. In this way, we can predict their future travel plans and help their travels, showing the power of data mining urban data in helping to solve future urban problems as well as diagnose and improve existing problems. Thesis Supervisor: Kent Larson Title: Principal Research Scientist in MIT Media Lab Thesis Supervisor: Takehiko Nagakura Title: Associate Professor of Design and Computation in Architecture Department Reader: Michael Dennis Title: Professor of Architecture Department