Filaments of crime: Informing policing via thresholded ridge estimation

摘要

In this study, we investigate the potential for optimizing hot spot patrol routes through density ridge estimation. We explore the application of an extended version of the subspace-constrained mean shift algorithm by using 2018 and 2019 Part I crime data from Chicago. Ultimately, the goal of mapping hot spots is to show concentrations of crime, thus targeting the epicenters only focuses on one problem area. For this reason, we refine patrol optimization to focus on the critical ridges in hot spots. In doing so, we extract density ridges of 2018 to early 2019 Part I crime incidents from Chicago to demonstrate that nonlinear mode-following ridges agree with broader kernel density estimations. We create multi-run confidence intervals and show that our patrol templates cover around 94% of incidents for 0.1-mile envelopes around ridges, and deliver evidence that ridges following crime densities enhances the efficiency of patrols. Our post-hoc tests show the stability of ridges, thus offering an alternative patrol route option that is effective and efficient.