Weighted-traffic-network–based geographic profiling for serial crime location prediction

Abstract

Geographic profiling plays a significant role in serial crime detection nowadays, in which Rossmo's formula is applied for future crime location prediction. However, the limited accuracy and demanding for vast data have largely impeded the efficiency of this technology. In this letter, a traffic network is introduced to geographic profiling. The problem is remodeled with weighted traffic network and the original Euclidean distance is replaced by the shortest path between nodes for better location prediction. A serial crime case is used to validate the correctness, efficiency and robustness of the proposed method. The main contributions of this letter can be concluded as follows: 1) the proposed model displays a higher accuracy and is less dependent on crime data; 2) strong robustness is testified by sensitive analysis, i.e. the developed model can produce an accurate prediction based on somewhat inaccurate former crime data; 3) further application in counter-terrorism is put forward with some adjustments.