Improving Classification Efficiency of Orthogonal Defect Classification via a Bayesian Network Approach

Orthogonal defect classification (ODC) is a kind of defect analysis method invented by IBM. ODC classifies software defects by eight orthogonal attributes. By analyzing these attributes´ distribution and increasing trend the software process information could be obtained. It has been used widely in many companies and organizations. In this paper, we focus on the ODC records collected in a company, and research to use these data to provide guidance in actual defect management to improve the efficiency of the classification. We study the relationships of these attributes and give a Bayesian network model, then with the help of the ODC records we got, a Bayesian network for ODC is presented. It shows great help in actual work for both the developers and the testers.