Bayesian Approach to Predict Offender’s Probable Anchor Point Using Geographic Profiling System in Akure, Nigeria

Abstract

One of the vital clues offenders leave behind when they commit crime is the crime location.  In recent years, several algorithms have been used to forecast the home site of unknown serial offenders on the basis of crime locations that have been linked to one offender. These have developed from spatial typologies to softwares that can provide direct support to crime investigations. Geographic profiling is an investigative methodology used in criminology that analyses the locations of a linked series of crimes to decide the most probable location for where the offender lives. This work establishes Bayesian Dirichlet Process Mixture Model (DPM) for crime hotspots analysis and as a geographic profiling system. It expands physiological profiling approach by integrating the idea of distance decay and buffer zone. The model was then implemented and tested with 119 spatial data of report serial theft cases in Akure, Nigeria. GPS Garmin was used to collect the data. The reported crime locations were visited to gather the data and were pre-processed by converting it into the machine readable format. The final output of the analysis (geoprofile) using the model was developed that depicts the most probable area of criminal(s) anchor point. A probability score was calculated for every point within the study area to indicate the likelihood that it contained the offender’s residence. The model was implemented in R. The model provides a practical tool for criminologist in targeting interventions and a more efficient use of resources for serial crime investigation. It can assist law enforcement agencies in decisions and policies making.