Mobility Data: New Ways to Measure Socio-economic Development & Predict Activity in Urban Spaces
CDS Moore-Sloan Data Science Fellow Anastasios Noulas focuses his research on urban change
The way people inhabit metropolitan landscapes changes over time, and it often changes rapidly. Traditional methods for tracking urban change — like censuses and surveys used by governments and special interest groups — can be costly, cumbersome, and out-paced by the rate of change in a metropolis.
Anastasios Noulas, a Moore-Sloan Data Science Fellow at NYU CDS, addresses this problem by developing new data-driven methods for tracking changes in urban populations and environments. His methods use digital datasets from online social media sources. Two of Noulas’s recent projects involve population activity in London.
In one project, Noulas and his collaborators used data analytics to determine the effects of cultural investment on socio-economic development. The researchers leveraged existing census data from 2010 and 2015, government spending reports about investment in cultural development, and population mobility data from user check-ins on the location-based social media platform Foursquare.
Through the application of machine learning algorithms and network metrics, the researchers performed an analysis on the relationship among these three data sets to map how cultural investment impacted specific London neighborhoods over time. They also defined a new set of metrics and proposed a highly accurate model to predict the effects of cultural investment. Their prediction model can help governments more effectively target resources to stimulate culture and commerce in disadvantaged communities.
In another project, Noulas and collaborators were able to predict the temporal visitation patterns of newly established businesses by using a similar dataset as input for machine learning models. The temporal visitation patterns identified by the machine learning model quantified what we intuitively know: people are more likely to visit certain types of venues (such as bars, gyms, restaurants, or commuter hubs), on certain days of the week, at certain times of day, in certain neighborhoods.
The prediction model from Noulas’s project, however, is far more useful than simply confirming common intuition. With a high degree of specificity, the algorithm can predict the number of visitors at any venue visited by Foursquare users. This prediction capability is especially useful for business owners who want to target the best resources for marketing, promoting, supplying, or maintaining a venue.
By Paul Oliver