The Use of Data Mining Techniques in Crime Trend Analysis and Offender Profiling

Abstract

The aim of this project is to ascertain whether the data in existing Police recording systems can be used by existing mature data mining techniques in an efficient manner to achieve results that are more accurate than those achieved by Police specialists when analysing crime. The Police Service has no formalised methodology of recording and analysing crime data and it is incumbent on each Force to train and develop appropriate personnel to provide operational analysis. Police data is inconsistent and, frequently, incomplete making the task of formal analysis far more difficult and current analytical practices are semi-manual and time consuming producing results of limited accuracy. These analytical processes would benefit from using data mining techniques within a structured approach as discussed within this thesis. The usage of supervised and unsupervised learning techniques within a structured methodology to mining Police data is evaluated. The research demonstrates that data mining techniques can be successfully used in operational policing. High volume crimes such as burglary that have been committed by one or more known offenders can be classified and the model used to attribute currently undetected crimes to one or more of those known offenders. Burglary crimes that previously had no overt relationship and the identity of the offender is unknown can be clustered with the ability to suggest one or more offenders who may be responsible for committing the crime. The same techniques used in analysing high volume crime can be used to link low volume major crimes such as serious sexual assaults. The recognised benefits include an improvement in the accuracy of results over current semi-manual processes and a reduction in the time taken to achieve those results.