Software quality classification using Bayesian belief networks

Bayesian Belief Networks (BBNs) have recently attracted attention as a possible solution for the problem of decision support under uncertainty. Although the underlying theory (Bayesian probability) is well founded, building and executing realistic models has only been made possible because of recent algorithms and software tools that implement them. There are no prior applications of BBNs to software quality classification models to our knowledge. In this paper, we investigate the application of BBXs to software quality classification modeling. We first give a brief overview of BBN modeling. Then, using software measurement data from very large C++ based software systems, we develop a BBN classification model. The independent variables are raw software measures collected from the system under test. Then we report the predictive quality of the BBN model and show that a BBN model can yield useful classification results.