Predicting C++ program quality by using Bayesian belief networks

There have been many attempts to build models for predicting the software quality. Such models are used to measure the quality of software systems. The key variables in these models are either size or complexity metrics. There are, however, serious statistical and theoretical difficulties with these approaches. By using Bayesian belief network, we can overcome some of the more serious problems by taking more quality factors, which have direct or indirect impact on the software quality. In this paper, we have suggested a model to predicting the computer program quality by using Bayesian belief network. We found that the implementation of all quality factors were not feasible. Therefore, we have selected 14 quality factors to be implemented on an average size of two C++ programs. The selection criteria were based on the reviewer's opinions. Each node on the given Bayesian believe network represents one quality factor. We have drawn the BBN for the two C++ programs considering 14 nodes. The BBN has been constructed. The model has been executed and the results have been discussed.