Predictive Analysis of Long-Term Risk Factors of Incarceration for At-Risk Youth
Embargo until
Date
2017-08
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The use of predictive analytics to prevent crimes, or predictive policing, is increasingly used by governments as part of criminal justice and law enforcement strategies today. Broadening the application of predictive analytics to encompass primary prevention strategies better informs government policy by leveraging the additional dimension of identifying risk factors that predispose youth to incarceration. To predict incarceration, a decision tree model with 72% accuracy was developed using data from the Prevention Program, a longitudinal intervention conducted on 900 first grade students in urban Baltimore Public Schools in 1993. A K-means clustering model was applied to the population to identify three archetype profiles: females, adapted males, and maladapted males. Overall classroom behavior, authority acceptance, and overall behavioral problem contributed most to the clusters. Results show that peer-rated and teacher-rated behavioral factors primarily influenced both predictive models. This indicates that peers and teachers are a valuable resource for identifying at-risk youth.
Description
Keywords
predictive analytics, machine learning, youth, prevention, incarceration