犯罪捜査実務における地理的プロファイリングの精度評価

抄録

Although previous studies have investigated whether the algorithm-driven technique is more accurate than human judges in geographic profiling (GP), those studies have used inappropriate methods in evaluating those accuracies. This study compared the algorithms (Geometric Region Methods: GRM, Spatial Distribution Strategies: SDS, and Probability Distance Strategies: PDS) in GP with the expert profilers’ judgments based on the 36 records of serial crimes. Most of the expert profilers’ analyses applied GRM such as circle hypothesis and suspicious area. Whereas the areas predicted by the expert profilers were significantly smaller than those of GRM, there was no significant difference between them in the rate of offenders’ home bases included in the predicted areas. Additionally, when comparing similarly-sized areas calculated using SDS and PDS to the areas predicted by the profilers, more offenders’ home bases were included in the predicted areas in experts’ judgments than SDS and PDS. The reasoning used in experts’ judgments that formed the basis of their accurate and efficient predictions and the practical use of PDS were discussed.