Exploring women's night walking safety in London from built environment perspective

Adopting a deep learning approach to provide a safer built environment for women walking at night.

In response to the Mayor of London's Charter for Women's Safety at Night, this project aimed to better understand how the built environment features relate to crimes that occur in women night’s walking and perceptions of risks. To do so, the project conducted a review of the literature on a) the relationship between built environment and crime, b) the relationship between built environment and perception of walking safety, and c) the perception difference in walking safety between women and men. 

The project also undertook data collection, incorporating: Twitter data, crime data, built environment data, perception of safety data. This was then analysed through: 

• analysed the Google Street View images through a deep learning method, image segmentation. Specifically, using the deep learning to extract the proportion of six elements of Google Street View images, which are sky, vehicles, buildings, vegetations, streetlights, road and footpaths
• used binomial regression model to examine the linkage of built environment characteristics and crime rates
• explore the differences in risk perception of night walking between women and men: this research used thematic analysis, sentiment modelling, topic modelling to analyse the data from the fieldwork diary (non-participant observation) and the scraped twitter data. 

The research collaboration also enhanced interdisciplinary planning to provide a safer built environment for women walking at night, in particular by extracting built environment characteristics from big data to understand the relationship between crime and the built environment. The project team now plan to publish a paper based on the data collected and preliminary results. This project will act as a seed fund to help apply for the next grant to conduct a more comprehensive comparative study between European cities. 

Project leads also noted that, "the funding provided a valuable chance for us to collaborate together and learn from each other. The PhD students in the department also benefitted to some extent. Specifically, one PhD candidate got the chance to be involved in the research and learned how to conduct text mining." 

Image credit: Photo by Filip Mroz on Unsplash