Publication: IEEE Transactions on Software EngineeringAuthors: Yasutaka Kamei; Emad Shihab; Bram Adams; Ahmed E. Hassan; Audris Mockus; Anand Sinha; Naoyasu Ubayashi
Summary
This paper aims to develop a defect prediction model at the change (i.e. commit) level. They gathered 14 important change measures mentioned in previous works and built a defect prediction model based on that. They utilized 5 open source systems and 6 commercial systems as their dataset. SZZ algorithm was used to identify the defect inducing change for the OSS projects. For the commercial projects, they utilized a root-cause analysis database provided by the vendors to identify mark the defect inducing change. As the prediction model, they used logistic regression.
Contribution
Gathered a set of 14 important change measures based on the previous works
Built a logistic regression model that can predict defect-inducing changes with 68 percent accuracy and 64 percent recall
Showed that defect prediction at change level requires only 20% efforts with respect to prediction at the module level
Identified the major characteristics of defect inducing changes
Publication: IEEE Transactions on Software Engineering Authors: Yasutaka Kamei; Emad Shihab; Bram Adams; Ahmed E. Hassan; Audris Mockus; Anand Sinha; Naoyasu Ubayashi
Summary
This paper aims to develop a defect prediction model at the change (i.e. commit) level. They gathered 14 important change measures mentioned in previous works and built a defect prediction model based on that. They utilized 5 open source systems and 6 commercial systems as their dataset. SZZ algorithm was used to identify the defect inducing change for the OSS projects. For the commercial projects, they utilized a root-cause analysis database provided by the vendors to identify mark the defect inducing change. As the prediction model, they used logistic regression.
Contribution