Software Defect Prediction in the Cloud

Abstract

Software defect prediction is one of the software quality assurance activities that can be applied during the software development life cycle. This activity helps quality assurance groups and project managers to determine risky modules which require more attention and more testing efforts. In this study, we investigated nine classification algorithms on 22 datasets which contain class-level metrics as part of our withinproject case study. After this case study, we sorted the datasets regarding their data instances and performed cross-project experiments on the large datasets, namely Apache Xalan, Xerces, and POI projects. We demonstrated that Decision Tree based algorithms are mostly superior to the other classification algorithms for within-project defect prediction and acceptable results can be achieved with Logistic Regression algorithms for cross-project defect prediction even if the data transformation approaches are not applied.

Keywords

• Data mining
• Machine learning
• Data management
• Software quality engineering
• Software defect prediction
• Cross-project defect prediction
• Cloud computing