Connecting the dots: assessing the accuracy of geographic profiling software

Abstract

Purpose

The purpose of this paper is to conduct an independent analysis of all existing geographic profiling software packages to determine if any one is more accurate than the others or if any of the software systems are any more accurate than simple spatial distribution strategies at locating the home base of serial offenders.

Design/methodology/approach

An analysis was conducted of all existing geographic profiling software as well as three spatial distribution methods of profiling. Differences in accuracy were assessed using four different methods; dichotomous profile accuracy, simple error measurement, profile error distance, and average top profile area.

Findings

Results indicate that not only are the different profiling software systems no more accurate than the spatial distribution control methods, but that accuracy in general was marginal at best. In addition results indicated that certain crimes, such as commercial robbery, were particularly difficult to profile and that the number of crimes in a series was not by itself a good indicator of success of a profile.

Research limitations/implications

The paper shows that future research needs to focus more on determining how various factors such as city type, crime type, road network and spatial aspects of a crime series (dispersion and search area) impact profiling accuracy. In addition future research should also endeavor to determine whether these advanced strategies are substantially more accurate than other simple profiling strategies such as human prediction. Finally, future research should also seek to examine geographic profiling in a real world setting and how geographic profiling impacts the success of open investigations.

Practical implications

Practically, this study casts doubt not only on the overall accuracy of profiling strategies in predicting the likely home location of an offender, but also on whether probability strategies are substantially better than spatial distribution strategies.

Originality/value

This research was the first to independently analyze all of the existing geographic profiling systems against control methods for the purpose of determining the accuracy of these different methods.

Citation

Paulsen, D.J. (2006), "Connecting the dots: assessing the accuracy of geographic profiling software", Policing: An International Journal, Vol. 29 No. 2, pp. 306-334. https://doi.org/10.1108/13639510610667682