Advancing urban management: Integrating GIS, LLMs, and media narratives into environmental and socio-economic analyses for enhanced urban crime analysis

Abstract

This study examines how crime is reported in the media, how the public perceives crime, and how urban environmental and economic factors interact at both global and local spatial scales to shape criminal activity. The research analyzed 3,507 reports from the top 100 media outlets from 2023 to 2024 and utilized a large language model to extract key information. The test results showed that the inter-coder reliability between human labeling and LLM labeling datasets reached 0.92 and examining the discrepancies between media narratives and actual crime data. By applying traditional econometric and Geographic Information Systems (GIS) geospatial regression techniques, the study revealed the impacts of environmental and economic factors on crime distribution, as well as the spatial relationships between urban structural features (such as road network density and vacant buildings) and crime rates in the Chicago area. Additionally, through Geographically Weighted Regression (GWR), the study further investigated the local differences between crime rates and socio-economic and environmental factors. The research found that selective news reporting might distort public understanding and affect policy responses. The findings provide insights for urban safety strategies and crime prevention policies, emphasizing the need to enhance the accuracy of media reports and public education, and offer valuable guidance for policymakers and urban planners on how to improve urban safety through environmental and economic planning.