Recovering Commit Branch of Origin from GitHub Repositories

An approach to automatically recover the name of the branch where a given commit is originally made within a GitHub repository is presented and evaluated. This is a difficult task because in Git, the commit object does not store the name of the branch when it is created. Here this is termed the commit's branch of origin. Developers typically use branches in Git to group sets of changes that are related by task or concern. The approach recovers the branch of origin only within the scope of a single repository. The recovery process first uses Git's default merge commit messages and then examines the relationships between neighboring commits. The evaluation includes a simulation, an empirical examination of 40 repositories of open-source systems, and a manual verification. The evaluations show that the average accuracy exceeds 97% of all commits and the average precision exceeds 80%.