Application of Cluster Analysis in Research on the Spatial Dimension of Penalised Behaviour

Abstract

This paper is focused on some of the possibilities of the use of cluster analysis (clustering) in criminology and the sociology of law. Cluster analysis makes it possible to divide even a large dataset into a specified number of subsets in such a way that the resulting subsets are as homogenous as possible, and at the same time differ from each other substantially. When analysing geographical data, e.g. describing the location of crimes, the result of cluster analysis is a division of a territory into a certain number of coherent areas based on an objective criterion. The division of the territory under study into smaller parts is more insightful when the clustering method is applied compared to an arbitrary division into official administrative units. The paper provides a detailed description of hierarchical cluster analysis methods and an example of using the Ward’s hierarchical method and the k-means combinational method to divide data on crime reports in the city of Baltimore between 2014 and 2019. The analysis demonstrates that the resulting division differs considerably from the administrative division of Baltimore, and that increasing the number of groups emerging as a result of cluster analysis leads to an increase of variance of variables describing the structure of crime in individual parts of the city. The divisions obtained using clustering are used to verify the hypothesis on differences in crime structure in different areas of Baltimore. The main aim of the paper is to encourage the use of modern methods of data analysis in social sciences and to present the usefulness of cluster analysis in criminology and the sociology of law research.

Keywords

• cluster analysis
• environmental criminology
• geography of crime
• crime in baltimore
• computational social science