Classifying Adult Probationers by Forecasting Future Offending. Final Technical Report
| Cataloged on:
Jun. 25, 2012
ANNOTATION: If you are involved with the development and implementation of risk prediction tools in a probation setting, you should be aware of this new assessment system. “Random forest modeling techniques represent an improvement over the methodologies of traditional risk prediction instruments. Random forests allow for the inclusion of a large number of predictors, the use of a variety of different data sources, the expansion of assessments beyond binary outcomes, and taking the costs of different types of forecasting errors into account when constructing a new model. This study explores the application of random forest statistical learning techniques to a criminal risk forecasting system, which is now used to classify adult probationers by the level of risk they pose to the community” (p. 3). Sections of this report include: abstract; executive summary; introduction; background and history; unit of prediction and time horizon; outcomes forecasted by the models; errors and costs; three different live models; predictors used; trees, forests, and how the model functions; the influence of predictors on forecasted outcomes; forecasting accuracy; validation and comparison using the 2001 cohort; long term offending patterns in the 2001 cohort; steps needed to implement forecasting; and conclusion.