Geostatistical risk analysis of static and dynamic crime

The Geography of Crime has a history in criminology that repeatedly finds a clustering of crime in time and space. Research in this field explores spatio-temporal patterning by studying who commits crimes, and why and when they commit crimes more in some parts of a city. Research finds a level of stability for many crimes. This thesis uses reported assault and break-and-enter crimes in Regina to explore in more depth the spatial-temporal patterns of crime and to develop and use hazard-risk modelling to improve methods of predicting future crime concentrations. Specifically, in order to explore and improve current hazard-risk models in criminology, new software was created for this thesis (Crime Risk Assessment Software) in order to generate a geostatistical risk model of two crime types, assault and break-and-enter (dynamic and static, respectively), to determine whether geostatistics, specifically Kriging techniques, could create strong predictive risk surfaces of these crimes. Through this exploratory spatio-temporal research, it was believed that after buildling and testing the model, statistics would reveal that the Kriging model would more accurately predict static crime than it would dynamic crime owing to the mobility issue of the crimes chosen (e.g., assault can happen anywhere spatially whereas break-and-enter can only occur at a static location such as a residence). Using examples provided by the Regina Police Service (RPS) in Saskatchewan, Canada, from assault and break-and-enter data gathered over the period from January 1, 2005, to December 31, 2005, both the Kriging risk models and the Crime Risk Assessment software demonstrated successful application in depicting spatially clustered data consistent with Geography of Crime Geography of Crime research.