Prediction of serial perpetrator residence: Part II—Evaluation of prediction model accuracy

Abstract

AbstractA novel approach for geographic profiling was developed which assesses and integrates available information and evidence relevant to the case for the development of a geographic profile (Part I). The approach is flexible when lesser information is available, in the form of a centrographic model for when solely the victim abandonment or murder sites are known and a perpetrator trek model for instances where both encounter and abandonment sites are available. Eleven case studies were used to evaluate these models including a comprehensive application of the evidence driven model to the Yorkshire Ripper investigation from the view of the 1980 advisory team. The calculation of weights for and inclusion of factors in the prediction of perpetrator residence appears to be a viable method for geographic profiling. This method demonstrated the lowest average search area across all cases when compared to both centrographic spatial distribution strategies and the probability distance strategies implemented in software. Implications for casework include a reduction of resource use per serial incident such as manpower, time, and software expenses.