Do historical metrics and developers communication aid to predict change couplings?

Developers have contributed to open-source projects by forking the code and submitting pull requests. Once a pull request is submitted, interested parties can review the set of changes, discuss potential modifications, and even push additional commits if necessary. Mining artifacts that were committed together during history of pull-requests makes it possible to infer change couplings among these artifacts. Supported by the Conway´s Law, whom states that “organizations which design systems are constrained to produce designs which are copies of the communication structures of these organizations”, we hypothesize that social network analysis (SNA) is able to identify strong and weak change dependencies. In this paper, we used statistical models relying on centrality, ego, and structural holes metrics computed from communication networks to predict co-changes among files included in pull requests submitted to the Ruby on Rails project. To the best of our knowledge, this is the first study to employ SNA metrics to predict change dependencies from Github projects.