Predicting the recurrence of domestic violence is easy: It will, most of the time. This research attempts to determine subsequent severity and to model a violent domestic setting. Co-investigators for this research are Dr. Christine Rasche (Department of Sociology, Anthropology, and Criminal Justice) and Dr. Susan Wallace (Department of Computer and Information Sciences).

Pilot work has been completed for this project. Traditional statistics and artificial neural networks were compared in terms of accurately predicting the severity of domestic violence. Linear regression techniques (nonlinear techniques will be used later) were not able to accurately identify any of the test cases or any of the training cases. The artificial neural network model was able to correctly identify all test cases with no false positives. This research was presented at the Annual Meeting of the American Society of Criminology in November, 1995. A journal article is under revision.

The current phase of the research is to examine more powerful statistical modeling techniques and compare them to artificial neural network models. Funding proposals are being developed.

Since July 30, 1998, you are visitor number to this page.