Statistical defect prediction models for software quality assurance

Software defects entail a highly-significant cost penalty in lost productivity and post-release maintenance. Early defect prevention and removal techniques can substantially enhance the profit realized on software products. The motivation for software quality improvement is most often expressed in terms of increased customer satisfaction with higher product quality, or more generally, as a need to position SAP Inc as a leader in quality software development. Thus, knowledge about how many defects to expect in a software product at any given stage during its development process is a very valuable asset. The great challenge, however, is to devise efficient and reliable prediction models for software defects. The first problem addressed in this thesis is software reliability growth modeling. We introduce an anisotropic Laplace test statistic that not only takes into account the activity in the system but also the proportion of reliability growth within the model.;The major part of this thesis is devoted to statistical models that we have developed to predict software defects. We present a software defect prediction model using operating characteristic curves. The main idea behind our proposed technique is to use geometric insight in helping construct an efficient prediction method to reliably predict the number of failures at any given stage during the software development process. Our predictive approach uses the number of detected faults in the testing phase. Data from actual SAP projects is used to illustrate the much improved performance of the proposed method in comparison with existing prediction approaches.