ABSTRACT
In recent years, a variety of algorithms have been used to predict the home location of unknown serial offenders on the basis of crime locations that have been linked to one offender (Canter and Larkin 1993; Rossmo 1993). These have evolved from spatial typologies such as the marauder/commuter distinction (Canter and Gregory 1994) to software packages that can provide direct support to investigations, often known as geographic profiling systems (Canter, Coffey, Huntley, and Missen 2000; Levine & Associates 2000; Rossmo 1993, 2000). Until very recently, however, the predictive accuracy of these geographic profiling systems had not been compared to other predictive methods.
