Spatiotemporal analysis and predictive modelling of crime patterns using GIS and remote sensing techniques

Abstract

The rising crime rates have become a critical challenge for communities, causing businesses to close or relocate, leading to loss of life and property, and diverting taxpayer funds from development to crime control. Combating crime is a shared responsibility, not just that of law enforcement. However, police often focus on evidence collection for convictions rather than addressing the root causes of crime, allowing offenders to continue criminal activities. This study conducted a spatio-temporal analysis and predictive modelling of crime patterns in Masaka district using GIS and remote sensing techniques. A quantitative approach was employed, analysing 4,651 crimes reported from 2022 to 2024 across various police divisions. Satellite imagery was processed through loading bands into the software (ArcMap 10.8), band stacking, training samples, determining sample signature and finally through supervised classification to assess land use/land cover (LULC) features. The relationship between LULC features with crimes such as sexual abuse, theft, robbery, assault, and breakings was got from overlaying the two, LULC and a given crime type at a time. Crime hotspots and coldspots were identified by Kernel Density Estimation (KDE), and analysis of different hotspots and coldspots was done with Getis-Ord Gi*. Predictive models were created by a space- time cube pattern mining through aggregation of points. They revealed new, persistent, and sporadic hotspots. Findings showed an increase in crime from 1,225 cases in 2022 to 1,788 in 2024, with theft, assault, breakings, sexual abuse, and robbery being the most common. Urban areas, especially Nyendo-Mukungwe, Masaka city, Kimanya-Kabonera, and Masaka rural divisions, reported the highest crime counts. Spatial autocorrelation analysis (Moran’s I) indicated significant clustering for sexual abuse, theft, robbery, and assault, but not for breakings which showed dissimilar values were near each other. A strong correlation was found between crimes and LULC features, particularly built-up and agricultural/green areas with all the crimes in the study. Most hotspots were in urban police divisions, with predictive modelling showing some hotspots becoming sporadic or emerging as new high-crime areas. The study underscores the need for proactive crime-fighting strategies using GIS and remote sensing to identify crime attractors and optimize police resource deployment. Enhanced patrols and focused interventions in identified hotspots and emerging crime areas can improve crime prevention efforts in Masaka district