Analysis of crime convergence in the EU using machine learning methods

This thesis examines the development of crime in the European Union during the period 2013–2023, with a focus on the analysis of convergence among member states. The main objective is to assess whether crime levels are converging across countries using a combination of traditional statistical methods and machine learning techniques. The analysis is based on Eurostat data on crime and economic indicators, specifically GDP per capita. Sigma convergence is applied to evaluate changes in the variability of crime over time, while cluster analysis (K-means) is used to identify groups of countries with similar crime profiles. The study focuses on selected types of criminal activity, including intentional homicide, robbery, burglary of private residential premises, and drug-related offences. The results indicate that crime convergence in the European Union is not uniform and varies by type of offence. While property crime shows a tendency toward convergence, drug- related crime remains highly heterogeneous. In the case of violent crime, variability between countries is low, suggesting that significant convergence may have already occurred in earlier periods. Cluster analysis reveals the existence of relatively stable groups of countries that correspond to broader socio-economic and geographical patterns, particularly the distinction between Western and Eastern Europe. The findings suggest that convergence occurs primarily within these groups rather than uniformly across the entire European Union. This thesis contributes to a better understanding of crime dynamics in the EU and demonstrates the value of combining statistical methods with machine learning techniques in the analysis of complex social phenomena. It also highlights the importance of institutional and socio-economic factors in the interpretation of cross-country comparisons of crime.