SPORS: A Suspect Recommendation System Based on Offenders´ Reconstructed Spatial Profile

According to Crime Pattern Theory, individuals all have routine daily activities which require frequent travel between several nodes, with each used for various purposes, such as their home, work, or shopping location. As people move about, their familiarity with the spatial areas around, and in between, the nodes increases, eventually forming their Activity Space. Offenders have similar spatial movement patterns and Activity Spaces as non-offenders, hence, according to theory, an offender will commit the crimes in their own Activity Space. Previous work in this research area determined the Activity Nodes in a city using clustering techniques based on the directionality of crime locations of repeat offenders. This paper extends that research by proposing a top-k recommendation system, called SPORS, which reconstructs the entire Activity Space for offenders and, for any new crime, recommends the top-k likely suspects for that crime. This algorithm is evaluated using information about 322 repeat offenders within the City of Surrey. Using only the spatial location of previous crimes and the home location of the offenders, SPORS is found to accurately predict the correct offender 22.30% of the time when predicting the top-10 suspects, a 718% improvement over naïve random selection.