Output details

DConfusion: A technique to allow cross study performance evaluation of fault prediction studies.

<09> This paper delivers a methodology for validating the results of binary classification studies, and also provides a tool for comparing different studies. The technique developed applies to assessing the performance of machine learning, not only in defect prediction, but any situation where a confusion matrix is produced. This paper is a source of corrigenda for published works, and addresses problems that have been reported in some highly cited papers (such as Elish and Elish 'Predicting defect-prone software modules using support vector machines' 2008). A preliminary version of this paper won the best conference paper at PROMISE 2012.