Using Predictive Analytics and Machine Learning to Improve the Accuracy and Performance of Juvenile Justice Risk Assessment Instruments: The Florida Case Study

• In Event: Risk Assessment:  Machine Learning vs. Regression—Which Direction Should We Go

Abstract

This paper presents the findings from a study designed to explore the potential for predictive analytics and machine learning to improve the performance of juvenile justice risk assessment instruments. More specifically, the research presented is a case study of the PACT (POSITIVE ACHIEVEMENT CHANGE TOOL). The PACT is the juvenile justice risk assessment model used by the Florida Department of Juvenile Justice (DJJ). The researchers were interested in exploring the potential for improving the accuracy and overall performance of the PACT and the possibilities for minimizing racial bias in the PACT in order to minimize the adverse consequences for youth of color. The findings from the study indicate that predictive analytics and machine learning (ML) increased the overall performance and accuracy of the PACT significantly. The authors believe that the main reason why such robust results were achieved is largely due to the fact that analytics and machine learning do a more effective job measuring the interactional effects between predictive and explanatory variables. The decision trees presented in the paper help to illustrate this. The paper describes the steps taken to prepare the data for analysis and build the predictive analytics and machine learning models and algorithms. Finally, the implications for policy and future research are discussed.

Author

• Ira Schwartz, Consultant and Advisor to Algorhythm