Abstract:In order to quickly predict and spot possible crime scene area, a geographic portrait method which is based on the evaluation of the residential places of a serial criminal was worked out. First, based on the crime scenes of a serial crime, a simplified CGT model is employed to spot the criminal’s domicile; then by using the concept of the maximum likelihood esti- mate Bayesian distribution, a multiple-dimension model which incorporates geographic bias and population density is construc- ted so as to predict the crime site. Lastly the efficacy of this model is proved via real criminal cases.