Risk identification for defect reduction in software development
Jeff Tian, A. Güneş Koru · 2004
This study is in the area of identifying the high risk software modules, which are defect prone or volatile (frequently changed) modules, by using complexity metrics. Such knowledge is valuable for software managers and developers since it can enable them to focus quality assurance activities on those modules. Using the complexity and defect data that belong to six large scale software products of our industrial partners, Nortel Networks and IBM, we found evidence that the most defective modules are not the same as the most complex modules, but the ones that fall between 65 and 85 percentiles in the complexity ranking. After our work on the closed source products, we performed research on the usability of defect data stored in the open source project repositories. Following that, we measured the complexity of two large scale software products, Mozilla and OpenOffice, and analyzed the complexity and associated change data. The open source results verified our previous findings. Our study points to the possibility of some high risk modules which immediately follow the most complex ones in the complexity ranking. These results should help software managers and developers improve the quality of their products by also saving valuable resources.