Predicting software revision outcomes on GitHub using structural holes theory - Archive ouverte HAL Access content directly
Journal Articles Computer Networks Year : 2017

Predicting software revision outcomes on GitHub using structural holes theory


Many software repositories are hosted publicly online via social platforms. Online users contribute to the software projects not only by providing feedback and suggestions, but also by submitting revisions to improve the software quality. This study takes a close look at revisions and examines the impact of social media networks on the revision outcome. A novel approach with a mix of different research methods (e.g., ego‐centric social network analysis, structural holes theory and survival analysis) is used to build a comprehensible model to predict the revision outcome. The predictive performance is validated using real life datasets obtained from GitHub, the social coding website, which contains 32,962 pull requests to submit revisions, 20,399 distinctive software project repositories, and a social network of 234,322 users. Good predictive performance has been achieved with an average AUC of 0.84. The results suggest that a repository host's position in the ego network plays an important role in determining the duration before a revision is accepted. Specifically, hosts that are positioned in between densely connected social groups are likely to respond more quickly to accept the revisions. The study demonstrates that online social networks are vital to software development and advances the understanding of collaboration in software development research. The proposed method can be applied to support decision making in software development to forecast revision duration. The result also has several implications for managing project collaboration using social media.

Dates and versions

hal-01667387 , version 1 (19-12-2017)



Libo Li, Frank Goethals, Bart Baesens, Monique Snoeck. Predicting software revision outcomes on GitHub using structural holes theory. Computer Networks, 2017, 114, pp.114--124. ⟨10.1016/j.comnet.2016.08.024⟩. ⟨hal-01667387⟩
79 View
0 Download



Gmail Facebook X LinkedIn More