Geographical profiling incorporating neighbourhood-level factors using spatial interaction modelling

The aim of this study was to examine the applicability of methods that use spatial interaction modelling to predict the most probable area for an offender's residence. Tokyo, which is the capital of Japan and is divided into 1507 square-kilometre zones, was selected as the study area. We analysed 4316 criminal trips to commit residential burglaries by 1089 offenders who lived in Tokyo. The following neighbourhood-level factors, and a distance-decay effect between zones, were incorporated into the proposed model: the size of the residential population aged 15-59 years in origin zones, the numbers of households living in detached houses, newcomers, the number of police facilities in destination zones, and the spatial structures of the criminal trips. Search areas calculated by the proposed model were smaller than previous models, suggesting that neighbourhood-level factors are important for predicting the location of an offender's residence.