Poster session presented at the annual conference of the American Psychology-Law Society in March 2009

By Wesley J. English, Evan Harrington, Nancy Zarse and David Canter

Abstract

Existing geographic profiling software is either too cost-prohibitive to purchase and implement or too difficult and time consuming for everyday use by law enforcement. A new geographic profiling software system was designed to overcome these limitations. The current study documents the development of GeoProfile and tests its reliability against two existing systems. A set of 55 serial offenders from Baltimore County, MD, were analyzed with GeoProfile, CrimeStat, and Dragnet using search cost and error distance as accuracy measures. Non-inferiority testing supported the hypothesis that GeoProfile is as accurate as CrimeStat and Dragnet.

Introduction

Problem

Geographic profiling researchers have created predictive software to help law enforcement locate serial offenders from their distribution of crime scenes. However, law enforcement has yet to widely embrace geographic profiling (Paulsen, 2006).

After reviewing the three major geographic profiling software systems, Rigel Analyst, CrimeStat, and Dragnet, it became clear that each exhibits one of two limitations: either the application is (a) too cost-prohibitive to purchase and implement or (b) too difficult and time consuming for everyday use by law enforcement.

Solution

A new geographic profiling software system, GeoProfile, was designed and developed to address these issues.

Hypothesis

It was predicted that GeoProfile would perform as accurately as existing geographic profiling software.

Method

Data

The data used to test the reliability of GeoProfile were crime scene and home locations of 55 serial offenders from Baltimore County, Maryland.

Apparatus

Comparing GeoProfile to all three of the major geographic profiling applications would have produced a fuller picture of where GeoProfile ranks among the others in terms of accuracy. However, D. Kim Rossmo, the software’s creator, denied a request to use Rigel Analyst in this study. Consequently, this study compares GeoProfile to CrimeStat 3.1 and Dragnet-K.

Procedure

The data set was analyzed by CrimeStat, Dragnet, and GeoProfile. Two measures of accuracy, search cost and error distance, were then measured and recorded for each series on all three applications. Search cost is defined as the percentage of the map searched before finding the actual home location of the offender (Rich & Shively, 2004). Error distance is defined as the straight-line distance from the predicted home location to the actual home location of the offender (Rich & Shively, 2004).

The scores for GeoProfile were then compared to the scores of CrimeStat and Dragnet using non-inferiority testing with the upper equivalence range defined a priori as 20% worse than the reference mean.

Results

Table 1: Non-Inferiority Tests
Accuracy Measure Application Non-Inferior P-Value
Search Cost CrimeStat Yes .00001
Dragnet Yes .0023
Error Distance CrimeStat No .0556
Dragnet Yes .000001

Statistical Analysis

  • GeoProfile was non-inferior to CrimeStat (p = .00001) and Dragnet (p = .0023) when search cost was used as a measure of accuracy.
  • GeoProfile was non-inferior to Dragnet (p = .000001) when error distance was used as a measure of accuracy.
  • GeoProfile was not non-inferior to CrimeStat (p = .0556) when error distance was used as a measure of accuracy.
    • However, a significant difference between the mean error distances of GeoProfile and CrimeStat was not found (p = .3372).
    • In this instance, even the established applications were not equivalent. Dragnet was also not non-inferior to CrimeStat (p = .6332) when error distance was used as a measure of accuracy. Furthermore, a significant difference between the mean error distances of Dragnet and CrimeStat was found (p = .0022).

Visual Analysis

  • Graphs of the software’s performance illustrate that GeoProfile follows the same pattern of performance at each increment of the data set as CrimeStat and Dragnet on both sets of accuracy measures.
Figure 2

Figure 2: Cumulative Search Cost

Figure 3

Figure 3: Cumulative Error Distance

Figure 4

Figure 4: Scaled Cumulative Search Cost

Discussion

Conclusion

GeoProfile was shown to be non-inferior to at least one of the major geographic profiling applications in each category of accuracy measures. Consequently, the hypothesis that GeoProfile performs as accurately as the established software systems is supported by these results.

Limitations

Including Rigel Analyst in the study would have provided a clearer picture of where GeoProfile stands in terms of accuracy in relation to all three of the major geographic profiling applications.

Additionally, a larger sample that included series from multiple jurisdictions around the country would have strengthened the results by eliminating any confounding variables related to Baltimore County.
Implications

The introduction of GeoProfile has two potential implications:

  1. GeoProfile could play an important role in advancing the field of geographic profiling by decreasing the amount of time and effort required to do research. Furthermore, GeoProfile allows another researcher the opportunity to experiment with the core strategies of geographic profiling, something that is limited to those with access to the inner workings of a program.
  2. GeoProfile could lead to the increased use of geographic profiling by law enforcement as the affordability and simplicity of GeoProfile may appeal to investigators who would be willing to implement geographic profiling if it were inexpensive and easy to use.

The Development of GeoProfile

Vision

To address the specific weaknesses of the existing geographic profiling software, GeoProfile was designed around four core concepts:

  • Simplicity
  • Speed
  • Affordability
  • Accuracy

Design

Simplicity: GeoProfile was designed as a web-based application so anyone with an internet connection could access the software regardless of their operating system. Moreover, no files need to be downloaded by the user; as a result, the software can be updated without having to redistribute the program to every user.

GeoProfile uses a simplified work flow to increase ease-of-use; it was designed to require little or no training. To create a profile, the user adds each crime scene by entering the latitude-longitude coordinates or street addresses, along with any desired meta-data, into a form. The predictive probability map is updated automatically with each change to the profile.

Furthermore, GeoProfile incorporates Open Street Maps, an alternative to Google Maps, so that a separate Geographic Information System (GIS) is not needed to apply the analysis to an investigation.

Speed: The simplified workflow and embedded mapping allow users to create and use a profile within minutes. One crime analyst reported creating a profile in eight minutes despite having never used GeoProfile before. Furthermore, users have the option of importing one or more series using a comma separated value (CSV) file.

Affordability: GeoProfile remains affordable through the use of license-free open source software components and services. Consequently, the overhead is limited to web server costs and the time it takes to maintain the server and software.

Accuracy: Because the core algorithm used by GeoProfile is the same as the one used by CrimeStat and Dragnet, GeoProfile is expected to preform with similar accuracy; this study tests that hypothesis.

Implementation

After the design for GeoProfile was established, Jim English, an applications developer by profession, developed a prototype of the application to test the study’s hypothesis. Upon completion of the study, Wesley English developed a stable version of GeoProfile for use by the law enforcement and research communities. The core algorithms remained the same as the prototype, but the user interface was redesigned to increase the usability and aesthetic quality of the software. Furthermore, the stable version implemented two distinct interfaces, one for law enforcement and one for researchers.

Special Thanks

Wesley English would like to thank the following people: Evan Harrington for his guidance as thesis chair; Nancy Zarse for her feedback as second reader; Jim English for programming the prototype; David Canter for providing Dragnet; and Ned Levine for providing CrimeStat and the data set.

References

  • Paulsen, D. (2006). Connecting the dots: Assessing the accuracy of geographic profiling software. Policing: An International Journal of Police Strategies and Management, 29, 306-334.
  • Rich, T., & Shively, M. (2004). A methodology for evaluating geographic profiling software. Cambridge, MA: Abt Associates.