Assessing the predictive performance of machine learners in software defect prediction

We present a methodology for Bayesian analysis of software quality. We cast our research in the broader context ofconstructing a causal framework that can include process, product, and other diverse sources of information regarding fault introductionduring the software development process. In this paper, we discuss the aspect of relating internal product metrics to external qualitymetrics. Specifically, we build a Bayesian network (BN) model to relate object-oriented software metrics to software fault content and faultproneness. Assuming that the relationship can be described as a generalized linear model, we derive parametric functional forms for thetarget node conditional distributions in the BN. These functional forms are shown to be able to represent linear, Poisson, and binomiallogistic regression. The models are empirically evaluated using a public domain data set from a software subsystem. The results showthat our approach produces statistically significant estimations and that our overall modeling method performs no worse than existingtechniques.