Mike O'Leary

Navigation Home Courses Publications Presentations Geographic Profiling Cyber Security Team Vita Contact

Links The School of Emerging Technologies CyberOperations Blog Department of Mathematics Department of Computer and Information Sciences

Introduction The geographic profiling problem is the problem of constructing an estimate for the location of the anchor point of a serial offender from the locations of the offender's crime sites.

We have developed a new mathematical algorithm for the geographic profiling problem based on Bayesian methods that allows us to allow for geographic features that affect crime site selection and for geographic features that affect the choice of an offender's anchor point.

The latest version was released in June 2012. Changes from the previous version include:

A new graphical user interface including a better help system
Support for Shapefiles for both input and output
Support for 2010 US Census data
The ability to use the locations of known past offenders to develop the prior distribution or the location of the series offender
The ability to manually specify the bandwidth when determining prior distributions, both of offenders and of crime locations
Correcting a round-off error in the previous version that could manifest itself at points far away from all crime sites

Testing has shown this tool is significantly more accurate than earlier versions. In a test of 237 solved Baltimore County residential burglaries, the prototype's search are contained the offender's home 70% of the time, including 47% of commuters.

Similar tests performed on 83 solved non-residential burglaries in Baltimore County showed that the prototype's search area contained the offender's home 74% of the time, including 66% of commuters.

This increase in accuracy however, comes with two caveats- the tool now takes significantly longer to run; it can take many hours to complete its analysis. It is also the case that the search areas are larger than those produced by previous versions of the tool.

Current version (.zip), Windows, Summer 2012.

I am happy to share the source code for the tool; please e-mail me and ask. The prototype remains under active development!

A complete description of the tool, how to use it, and the underlying mathematical model are all contained in the following technical report submitted to NIJ.

This work has been supported by the National Institute of Justice, through grants 2009-SQ-B9-K014, 2007-DE-BX-K005, and 2005-IJ-CX-K036.

We also wish to thank Baltimore County Police Department for providing some of the data used to develop and test these algorithms.

Related Presentations

Models for Offender Target Location Selection with Explicit Dependency Structures, Quantitative Methods in Defense and National Security 2012, Fairfax, VA April 30 - May 1, 2012. (.pdf)
Patterns in Offender Distance Decay and the Geographic Profiling Problem, MAPS Conference, Miami, FL, October 2011. (.pdf)
A New Software Prototype for Geographic Profiling , MAPS Conference, Miami, FL, October 2011. (.pdf)
Applying Mathematics.... to catch criminals Randolph Macon College, September 2011.
Patterns in Offender Distance Decay and the Geographic Profiling Problem, AMS-MAA Joint Meetings, New Orlans, LA, January 2011.
Patterns in Offender Distance Decay and the Geographic Profiling Problem, 2010 Fall Western Section Meeting of the AMS, Los Angeles, CA, October 2010. (.pdf)
Implementing a Bayesian Approach for the Geographic Profiling Problem, The First International Conference on Computing for Geospatial Research and Applications (COM.Geo 2010) Washington, DC, June 2010. (.pdf)
New Approaches for the Geographic Profiling Problem, The NIJ Conference, Arlington VA, June 2010 (.pdf)
Modeling Distance Decay for the Geographic Profiling Problem, Geospatial Technical Working Group, Portland OR, April 2010. (.pdf)
Mathematical models for geographic profiling problem, UCLA Applied Mathematics Colloquium, November 2009 (.pdf)
A new software tool for geographic profiling problem, The Tenth Crime Mapping Conference, New Orleans LA, August 2009
A new software tool for geographic profiling, The NIJ Conference, Arlington VA, June 2009
Mathematical models for the geographic profiling problem, Center for Evidence-Based Crime Policy, George Mason University, March 2009 (.pdf)
Mathematical models for the geographic profiling problem, Georgetown University Mathematics Department Colloquium, March 2009 (.pdf)
Determining the Optimal Search Area for a Serial Criminal, Joint Mathematics Meetings, Washington DC, January 2009 (.pdf)
Advances in Geographic Profiling 31^st Applied Geography Conference, Wilmington DE, October 2008 (.pdf)
Determining the Optimal Search Area for a Serial Criminal INFORMS National Meeting, Washington DC, October 2008 (.pdf)
Using Mathematics to Catch Criminal Stevenson University Kappa Mu Epsilon Meeting, September 2008 (.pdf)
The Mathematics of Geographic Profiling, The NIJ Conference, Arlington VA, July 2008 (.pdf)
The Mathematics of Geographic Profiling, Spring Meeting, Geospatial Technology Working Group, New Orleans, April 2008. (.pdf)
The Mathematics of Geographic Profiling, Center for Army Analysis, Fort Belvoir, April 2008 (.pdf)
The Mathematics of Geographic Profiling, Baltimore County Police Department, Towson MD, June 2007 (.pdf)
The Mathematics of Geographic Profiling, Ninth Crime Mapping Research Conference, Pittsburgh PA, March 28-31 2007 (.pdf) (Conference)
The Mathematics of Geographic Profiling, Crime Hot Spots: Behavioral, Computational and Mathematical Models, Institute for Pure and Applied Mathematics (IPAM), UCLA, January 29 - February 2, 2007. (.pdf) (Conference)
A New Mathematical Technique for Geographic Profiling, The NIJ Conference, Washington D.C., July 2006. (.odp) (.pdf)
Determining the Optimal Search Area for a Serial Criminal, Modeling and Simulation Technical Working Group, National Institute of Justice, Jersey City, March 2006. (.pdf)