Using data to build a safer city
- By Kathleen Hickey
- Oct 31, 2016
Urban planners soon may be able create safer cities based on a combination of cellphone data, Google Street View images and machine learning.
For several years, MIT researchers have been assigning streetscapes from Google Street View scores for safety, activity and affluence.
Safety scores initially were done by volunteers, but because of the vast number of images in the researchers’ database, three years ago the team began training a machine-learning system to assign scores to the remaining images. (Volunteers by then had provided more than 1.4 million comparisons.)
Scores were provided by a convolutional neural network trained on a dataset of Google Street View, according to MIT News. “That’s ultimately how you’re able to take this type of research to scale,” MIT professor César Hidalgo said. “You can never scale by crowdsourcing, simply because you’d have to have all of the Internet clicking on images for you.”
In a new study, researchers from MIT Media Lab, together with colleagues at Italy’s University of Trento and the Bruno Kessler Foundation, detailed how adding data from cellphone usage to the machine-learning/Google Street View project could be used to create safer cities.
Images from neighborhoods in Rome and Milan were accorded safety scores, and these scores were compared against the frequency with which people visited these places. An area’s popularity was determined by mobile phone billing, the researchers wrote, which the researchers used as a proxy for activity. For each grid cell, they counted the number of people who made or received a call on an hourly basis, broken down by gender and age.
They also built an algorithm that recorded changes to the safety scores assigned by the machine-learning system for specific areas of the city. The study controlled for population, employment density and distance to the city center.
Not surprisingly, characteristics perceived as creating a safe environment included street-facing windows and well-kept green spaces (poorly maintained green spaces were not perceived as safe).
Researchers did, however, find a strong correlation between the perceived safety of an area and the popularity of a location, particularly with women and people over 50. The correlation was negative for males in their 20s, who were more likely to visit neighborhoods generally perceived to be unsafe.
“Urban planning -- and there’s a lot of literature about it -- has been largely designed from a male perspective. ... This research gives scientific evidence that women have a specific perception of the appearance of safety in the city,” Luis Valenzuela, an urban planner and professor of design at Universidad Adolfo Ibáñez in Santiago, Chile, told MIT News.
Kathleen Hickey is a freelance writer for GCN.