Predicting fault-prone modules based on metrics transitions

Abstract

This paper describe a method for identifying fault-prone modules. The method utilizes metrics transitions rather than raw metrics values. Metrics transitions are measured from the source code of consecutive versions, which is archived in software repositories. Metrics transitions should be an good indicator of software quality because they reflect how the software system has evolved. This paper exhibits a case study, which is a comparison between metrics transitions and CK metrics suite. In the case study, the metrics transitions could precisely identify fault-prone modules.