Predicting high-harm offending using machine learning: an application to outlaw motorcycle gangs

Abstract

Risk assessment tools are used widely in the criminal justice response to serious offenders. Despite growing recognition that certain outlaw motorcycle gang (OMCG) members and their clubs are likely to be involved in crime, particularly serious crime, this is not an area where risk assessment tools have been developed and validated. The nature of offending by OMCGs, and policing responses to OMCGs, requires a novel approach to risk assessment. This study uses machine learning methods to develop a risk assessment tool to predict recorded high-harm offending. Results are compared with those of a model predicting any recorded offending. The model predicted high-harm offending with a high degree of accuracy. Importantly, the tool appeared able to accurately identify offenders prior to the point of escalation. This has important implications for informing law enforcement responses.