Authors: Cui Yongxiang*,
Topics: Remote Sensing, Urban Geography
Keywords: Crime, Nighttime light data, Random Forest
Session Type: Paper
Start / End Time: 8:00 AM / 9:40 AM
Room: Studio 2, Marriott, 2nd Floor
Presentation File: No File Uploaded
Crimes, as one of serious public safety problems, have significantly negative impacts on human beings. It is crucial to analyze spatial and temporal patterns of crimes for the allocation of police forces and urban planning. Traditional methods for analyzing crimes usually focused on survey data such as housing value, poverty level, unemployment rate, population and other socio-economic features. As crimes are closely related to human activities and nighttime light data has been widely acknowledged as a useful source for describing human activities, this research made an attempt to explore the usefulness of nighttime light data in the spatial and temporal analysis of robbery and theft crimes. By integrating various datasets including nighttime light intensity, point of interest (POI), traffic information and socio-economic characteristics such as population and housing value, this research employs the Random Forest algorithm to estimate the occurrence of crimes. The results showed that nighttime light intensity is a significant estimator and the accuracy of the estimation reached 93% when the Random Forest algorithm was used. By quantitatively analyzing how nighttime light intensity could affect crime activities, this study can shed some lights on crime research and provide useful information for policy makers to ensure urban public safety.