Fault-Prone Source File Analysis Focusing on the Contribution Entropy in Open Source Development

Open source software (OSS) products have been widely used for information systems, and a successful quality management of OSS development has become one of key topics in the information technology world. Since the development and maintenance of an OSS product is driven by various developers, it would be worthy to focus on their contributions and the cooperative structure. This paper proposes to measure a developer's contribution to a source file as the cumulative lines of code that he/she has changed on the file, and to evaluate the balance of contributions to the file among different developers in a form of an entropy. Through an empirical study using data collected from 10 major OSS projects, the following findings are reported: (1) a source file which has been maintained by two or more developers (entropy>0) is about two times more likely to be faulty than a file which has been done by only a single developer (entropy=0); (2) when two or more developers have maintained a source file and their contributions are more imbalanced (a lower entropy), the source file is more fault-prone.