Comparing the accuracy of geographic profiling

Abstract

Studies of geographic profiling (GP) have generally investigated the efficacy of two categories of GP strategies for predicting an offender's base. These strategies can be classified as follows: (a) spatial distribution strategies, assessed by center of the circle hypothesis, mean center, median center, and the center of minimum distance, and (b) probability distance strategies, assessed by linear, negative exponential, logarithmic, and lognormal distributions. GP strategies were compared based on the data of 333 residential burglars who had committed at least three offenses in the Tohoku region during the years 2004-2013. Search area (total area that is searched before locating the offender's base) was utilized as an index for accuracy measure. The results demonstrated that probability distance strategies are more accurate than spatial distribution strategies. We conclude that this is because probability distance strategies captured crime patterns of residential burglars more precisely than spatial distribution strategies.