ASSESSMENT, EVALUATION AND MAPPING OF INDIVIDUALLY CALIBRATED JOURNEY-TO-CRIME CRIMINAL GEOGRAPHIC PROFILING MODELS

M. Leitner, T. Pal

Louisiana State University, Department of Geography and Anthropology, Baton Rouge, USA

mleitne@lsu.edu

 

Criminal geographic profiling (CGP) is a decision support tool used by law enforcement to make estimates about the likely location of a serial offenders haven, which, in most cases, is the offenders residence. Modern CGP models include Dragnet, RigelTM Criminal Geographic Targeting, Predator and the CrimeStat III journey-to-crime and crime travel demand modeling routines.

 

The overall objective of this research is to evaluate the usefulness of individually calibrated journey-to-crime (JTC) criminal geographic profiling (CGP) models.

 

The data set used in this research consists of a total of 247 serial crimes with both incident and home locations of the arrested offender known. It comprises of nine different crime types, including larceny (51), arson (4), auto theft (31), robbery [commercial (76), street (17), and mixed (15)], rape (1), and burglary [residential (51) and commercial (1)]. All offenders were arrested for three or more of the same crime type in Baltimore County, Maryland between 1994 and 1997. This dataset has been used previously to evaluate and compare the accuracy of different CGP models and software. However, previous research used the CrimeStat III default values, when calibrating the JTC CGP models. In contrast, this research will calibrate the JTC CGP models individually. This is the novel part of this research.

 

The JTC CGP method estimates the serial offenders haven based on a mathematically calibrated distance decay function defined for the observed travel patterns of the calibration group. Ideally, the calibration group includes many crimes from the same crime type and has been collected for the same study area as the test group. The test group includes all crime scenes associated with the same serial offense, for which a geographic profile is estimated. The five different distance decay functions that will be evaluated in this research include the linear, negative exponential, truncated negative exponential, normal and lognormal.

 

Preliminary results show that individually calibrated distance decay functions significantly increase the accuracy of JTC criminal geographic profiles compared to when a JTC CGP is calculated with the CrimeStat III default parameters. However, preliminary results comparing the accuracy of CGP models derived from the same individually calibrated distance decay functions with other CGP models and spatial distribution measures have been inconclusive.