342 Offender Project Report

The project builds upon the approach of the Hampshire Violence Reduction Unit regarding the importance of Adverse Childhood Experiences as an important dimension in violent crime offending. The project aimed to review datasets originating from four local authorities’ children and education services blended with relevant police data and analyse them using Machine Learning (ML) algorithms. The purpose was to assess whether these tools could be useful in terms of identifying links between specific variables from the data and particular types of knife-related offending. Due to practical and time constraints, however, the research team was only able to work with a small dataset from the relevant Portsmouth services without also being able to blend the relevant police data with the services datasets. In order to provide some additional insights, the police dataset was analysed separately. Multiple ML algorithms were used for the analysis, such as K-Nearest Neighbours, Naïve Bayes, Random Forest, Decision Tree and Apriori. Although the datasets used offered particular analytic challenges for the research team, the classification and association rules analyses still provide interesting and promising results. More specifically, the results highlight the potential of ML algorithms to generate well-performing prediction models with more complete and balanced datasets. Additionally, both classification with the Decision Tree algorithm and mainly association rules offer insights in terms of variables relating to the offenders’ educational and children services, and police data and their links with certain offence types. Finally, the report provides insights and recommendations for more efficacious and consistent data collection for future projects in order to increase the analytic potential of these ML algorithms and paves the way for developing related projects that can expand on the conclusions drawn and test the hypotheses provided.