Using Random Forest Risk Prediction in the Philadelphia Probation Department

• NCJ Number: 241346
• Author(s): Geoff Barnes; Jordan Hyatt
• Date Published: August 2012
• Length: 2 pages
• Agencies: NIJ
• Publication Type: Program/Project Description; Presentation; Interview; Conference Material; Research (Applied/Empirical)

Annotation

This video and its transcript present interviews with three stakeholders who provide information on the background and implementation of the Philadelphia Probation Department's use of a risk-prediction model that enables the department to focus its limited resources on probationers who pose the highest risk of reoffending.

Abstract

The prediction model collects information on probationers that the department largely already had available and predicts the likely conduct of each probationer for the first 2 years of his/her probation term. Based on the model's forecasting, probationers are assigned to supervision units based on their risk classification. The highest risk probationers are placed in a unit that provides the most intense supervision, and those probationers predicted to commit no new offenses or relatively minor offenses are placed in a unit that receives a decreased level of supervision. The prediction model was developed through the cooperative efforts of academic researchers and probation practitioners. It ensures that everyone entering probation receives the same objective risk assessment that facilitates a fair and cost-effective use of resources for each probationer.