Testing the utility of neural network models to predict history of arrest in batterers

In prisons, risk assessments are typically based on retrospective reports of factors known to be correlated with violence recidivism. Previous studies have used linear models that rely on variables that have been linked to past history of intimate partner violence (IPV) based on men's report only. The current study compares the non-linear neural network model to traditional linear models in predicting a history of arrest for any crime in men who self-report a history of IPV. In addition, models that include men's report only were compared to models that also include the victim's report. The neural network models were found to be superior to the linear models in their predictive power. Models that included victim report were superior to models that did not include victim report. These finding suggest that the prediction of violence recidivism may be enhanced through the use of neural network models and through models that include information gathered from victims.