The study of the correlation between software project and product attributes and its modules quality status (faulty or not) is the subject of several research papers in the software testing and maintenance fields. In this paper, a tool is built to change the values of software data sets' attributes and study the impact of this change on the modules' defect status. The goal is to find those specific attributes that highly correlate with the module defect attribute. An algorithm is developed to automatically predict the module defect status based on the values of the module attributes and based on their change from reference or initial values. For each attribute of those software projects, results can show when such attribute can be, if any, a major player in deciding the defect status of the project or a specific module. Results showed consistent, and in some cases better, results in comparison with most surveyed defect prediction algorithms. Results showed also that this can be a very powerful method to understand each attribute individual impact, if any, to the module quality status and how it can be improved.
International Journal of Software Engineering and its Applications
Alsmadi, Izzat M., "Measuring Defect Datasets Sensitivity to Attributes Variation" (2012). Computer Science Faculty Publications. 18.