Understanding Street Theft Hotspots Using Machine Learning and Google Street Image

Keywords: Crime hotspot, Convolution Neural Network (CNN), machine learning, environmental criminology, prediction
Abstract Type: Paper Abstract

Abstract

Environmental crime, image acquisition technologies, and machine learning approaches have all had significant effects on crime research. Environmental criminology is introduced by Brantingham and Brantingham (1981, 1984) that a broad concept with the idea that crime is closely associated with the environmental conditions of places. It can contribute to crime prevention by systematically explaining the connection between land use, traffic patterns, urban design and daily activities, as well as the movements of victims and offenders. Along with this theory, fine-scale environmental crime analysis is now also possible thanks to the progress of machine learning technologies and numerous image sources. This study's objectives are as follows: 1) to examine whether environmental criminological characteristics of crime hotspots can be predicted through Convolution Neural Network (CNN); 2) to pinpoint whether actual environmental aspects of crime hotspots can be identified and used as basic data for improving the surrounding environment of crime; 3) to serve as a foundation for further research into how criminal analysis may be carried out using Google Street View and CNN. From 2014 to 2019, theft incidents in Los Angeles, California, were utilized for training and testing a model, then found the predictability of crime hotspots.