Spatiotemporal crime prediction in Northern Ireland

The ability to accurately predict where and when a future crime will happen is a precious concept in policing; several techniques have been developed focused on this premise. Near repeat analysis, hotspot mapping, and the theory behind crime attractors and generators have been used to predict crime. This research examines the effectiveness of these techniques in a Northern Ireland context. It assesses these techniques independently and in combination to determine how effective they are at predicting the spatial and temporal dimensions of crime within a Northern Ireland study area. The research focuses not only on predicting traditional crime types, such as assault, burglary, criminal damage, robbery, theft, and vehicle theft, but also incorporates anti-social behaviour. In addition, this research develops existing techniques by incorporating variables related to weather conditions and seasonal patterns to determine how these can influence overall predictive accuracy. This research finds that all three techniques for predicting crime perform better than a random approach. A hotspot mapping based technique generates the highest predictive accuracy across all crime types when used independently. However, on merging the predictive output from the hotspot mapping technique with output from techniques based on near repeat and crime attractors and generators theory, predictive accuracy increases for two crime types (anti-social behaviour and burglary). Finally, this research presents a framework that can aid in short-term crime prediction. This framework provides an understanding of crime prediction techniques' effectiveness and how embedding weather and seasonal patterns can influence the prediction process.