Isolation of software defects: extracting knowledge with confidence

Software maintenance is one of the most time and effort-consuming phases of software development life cycle. Maintenance managers and personnel look for methods and tools supporting scheduling and performing different software maintenance tasks. To make such tools relevant, they should provide the maintainer/manager with some quantitative input useful for purposes of interpretation and understanding what factors influence maintenance efforts and activities. In this paper, a comprehensive multi-model prediction system is proposed. It dwells on evidence theory and a number of rule-based models independently developed using different methods. Application of evidence theory leads to determination of confidence levels of the generated rules as well as obtained predictions. The study is concerned with the effort needed for the removal of defects in existing software maintenance data. A multi-model prediction system is developed and prediction results are analyzed.