A simple crime hotspot forecasting algorithm

Robert Bogucki, Jan Kanty Milczek, Patryk Miziuła

Keywords: crime hotspot, spatiotemporal point process, grid search

TL;DR: We efficiently forecast crime hotspots with use of a simple spatio-temporal point process and grid search.

Abstract: Crime hotspot forecasting is an important part of crime prevention and reducing the delay between a 911 call and the physical intervention. Current developments in the field focus on enriching the historical data and sophisticated point process analysis methods with a fixed grid. In the paper we present a simple spatio-temporal point process allowing one to perform exhaustive (literal) grid searches. We then show that this approach can compete with more complex methods, as evidenced by the results on data collected by the Portland Bureau of Police. Finally, we discuss the advantages and potential implications of the new method.