Learning where to inspect: Location learning for crime prediction

Crime studies conclude that crime does not occur evenly across urban landscapes but concentrates in certain areas. Spatial crime analysis, primarily focuses on crime hotspots, areas with disproportionally higher crime density. Using Crime-Tracer, a personalized random walk based approach to spatial crime analysis and crime location prediction outside of hotspots, we propose here a probabilistic model of spatial behavior of known offenders within their activity space. Crime Pattern Theory states that offenders, rather than venture into unknown territory, frequently commit opportunistic crimes by taking advantage of opportunities they encounter in places they are most familiar with as part of their activity space. Our experiments on a large crime dataset show that CrimeTracer outperforms all other methods used for location recommendation we evaluate here.