A MACHINE LEARNING CLUSTER MODEL FOR THE DECISION-MAKING SUPPORT IN CRIMINAL JUSTICE

Abstract

In a modern digital society, information technologies play a crucial role in security policy. The increase in the number of criminals and the expansion of the range of crimes committed by them, which is observed all over the world, poses serious risks to the personal safety of citizens, the internal security of the country, and international security. Identifying links between the individual characteristics of prisoners and their criminal recidivism can help to solve serial crimes, develop new crime prevention strategies, and provide reliable support for public safety decisions.

The presented work is a part of research on the development of information and analytical support for decision-making systems in criminal justice. This document presents a new analytical approach to criminal profiling. It is a case study of a unique real-world dataset of 13,010 criminal convicts. The k-means clustering technique was used to determine significant indicators (individual characteristics of prisoners) that determine the propensity of convicts to commit repeated criminal offenses. The built clustering model makes obvious the connection between the propensity for criminal recidivism and the following elements of the criminal profile: the number of previous convictions, the age at the time of the first conviction, the presence of conditional convictions, and early releases. The developed models can be applied to new criminal convicted datasets. The dynamic interaction of information technology and the criminal justice system will help reduce crime and strengthen internal security.