Mining Software Repositories to Identify Library Experts

Abstract

Programming is multi-faceted, inherently involving several different skills. With the advent of collaboration platforms like GitHub, developers have the opportunity to contribute to projects from different organizations and collaborate with various developers from around the world. With GitHub data, new opportunities to identify developers abilities become possible. From GitHub, it is possible to infer several skills from a developer, for instance, the user of libraries. In this paper, we propose a method to identify library experts, based on the knowledge they produce on GitHub. We evaluated our method in an experiment to identify possible experts in three Java libraries. Our method ranked the top 100 developers for each technology. Then we compared the selected profiles from GitHub with profiles of these developers on the social network LinkedIn to see if what they report in LinkedIn matches what they produce in GitHub. We also surveyed students to compare the results of our method to the manual analysis. Our results showed that 89% of selected GitHub developers reported their skills in social networking sites as LinkedIn, according to the ranking made by our method and that the ranking produced by our method is related to the classification made by survey participants.