Authors: Alejandro Sanchez Zarate*, UN-Habitat
Topics: Digital Geographies, Latin America, Spatial Analysis & Modeling
Keywords: Digital Geography, Gentrification, Big Data, Mexico
Session Type: Virtual Guided Poster
Start / End Time: 1:30 PM / 2:45 PM
Room: Virtual 53
Presentation File: Download
The main purpose of this work is to explore the spatial pattern of principal topics in georeferenced tweets posted in gentrified neighbourhoods in Mexico City. It was made an assemblage of big data and official data. On one side, it was identified gentrified areas following a centrality of commercial establishments linked to gentrified consumption. On the other hand, it was harvested a database of 2 millions of georeferenced tweets in Mexico City from April 2017 to April 2019. Tweets were preprocessed to get plain text, delete stopwords and regular expressions. After that, tweets were temporally classified by weekday or weekend and, spatially if tweets were emitted in a gentrified area or not. Methodologically, it was used relative frequency of words to get the most repeated words by day and neighbourhood. In order to get the latent topics in tweets, it was applied a Latent Dirichlet Allocation Model (LDA), non-supervised Machine Learning algorithm. Principal findings point out to different lexicon between gentrified areas and not. Besides, 20 latent topics were recognised by LDA. These topics are related to gentrified practices, e.g., hipster o gay culture, food consumption, musical events and so on. Finally, tweets were classified by LDA and represented in hotspots maps. As a result, it was generated a visualisation of gentrified topics on Twitter in Mexico City.