Identifying irregularly shaped crime hot-spots using a multiobjective evolutionary algorithm |
| |
Authors: | Xiaolan Wu Tony H Grubesic |
| |
Institution: | (1) Department of Geography, Central Michigan University, 276 Dow Science Building, Mount Pleasant, MI 48859, USA;(2) Department of Geography, Indiana University, Student Building 120, Bloomington, IN 47405-7100, USA |
| |
Abstract: | Spatial cluster detection techniques are widely used in criminology, geography, epidemiology, and other fields. In particular,
spatial scan statistics are popular and efficient techniques for detecting areas of elevated crime or disease events. The
majority of spatial scan approaches attempt to delineate geographic zones by evaluating the significance of clusters using
likelihood ratio statistics tested with the Poisson distribution. While this can be effective, many scan statistics give preference
to circular clusters, diminishing their ability to identify elongated and/or irregular shaped clusters. Although adjusting
the shape of the scan window can mitigate some of these problems, both the significance of irregular clusters and their spatial
structure must be accounted for in a meaningful way. This paper utilizes a multiobjective evolutionary algorithm to find clusters
with maximum significance while quantitatively tracking their geographic structure. Crime data for the city of Cincinnati
are utilized to demonstrate the advantages of the new approach and highlight its benefits versus more traditional scan statistics. |
| |
Keywords: | |
本文献已被 SpringerLink 等数据库收录! |
|