Micro-built environment and crime: An approach using street view images and machine learning methods

Keywords: Place Perception, Micro-built environment
Abstract Type: Paper Abstract

Abstract

Crime events are threats to urban public security and residence life. The research found that crime events are related to the built environment characteristics. Studies often capture the built environment characteristics through conventional built environment data on a neighborhood scale, which lacks investigation of micro-built environment (MBE) characteristics. The ability to capture the MBE characteristics by street view images has been demonstrated by empirical research. This study associate reported street crime events and the MBE characteristics around the location. The MBE characteristics include physical objects and people's perceived perceptions extracted from crime locations by a deep learning model. Results show a significant association among physical objects, perception variables, and crime events. Mobility-related MBE, safety, and depressing perceptions have positive associations with crime events, while wealthy and beautiful perceptions have negative associations. This result further investigates the connection and impact of the surrounding built environment and crime.