Home      Login


Machine Learning Risk Assessment at Preliminary Arraignments for Domestic Violence  


Author:  Richard A. Berk.; Susan B. Sorenson.; Geoffrey Barnes.


Source: Volume 09, Number 01, Summer 2016 , pp.69-73(5)




Family & Intimate Partner Violence Quarterly

< previous article |next article > |return to table of contents

Abstract: 

The authors applied machine learning procedures to data from over 28,000 domestic violence arraignments in a major metropolitan area in order to determine which factors were most important in accurately predicting future violent offenses. “Machine learning” is a relatively new approach to risk assessment in the criminal justice field, but is finding more and more support because its use of algorithms, rather than models, has proved more reliable. This article summarizes the basic principles of machine learning as applied to the prediction of domestic violence.

Keywords: Actuarial risk assessment; “random forests” methodology;

Affiliations:  1: University of Pennsylvania; 3: Cambridge University.

Subscribers click here to open full text in PDF.
Non-subscribers click here to purchase this article. $15

< previous article |next article > |return to table of contents