Machine Learning and the City

To the first time, the availability of large amounts of geotagged user-generated content, e.g., in the form of social media feeds such as Tweets, enables us to understand the collective perceptions of different places. In this research, we applied Latent Dirichlet allocation (LDA) to aggregated tweets in each land parcels to extract the topics of different locations. Twenty topics out of 100 obtained from LDA were further labeled as location-related topics and corresponding land parcels were characterized using these topics. The most significant location-related topics such as ‘airport’ and ‘university’ are found near airport and university respectively.