FARIAS, A. M. M. A; http://lattes.cnpq.br/7471050021771055; FARIAS, Ariann Michael Martins de Andrade.
Resumo:
GitHub, as currently the biggest platform for hosting software development and version control, handles
on a daily basis, a massive stream of interactions between users and repositories. With millions of
repositories hosted on the platform, some projects that could be interesting for some users ended up
being unnoticed, same as other projects which are searching for developers ended up staying on a limbo.
In these situations, it becomes obvious the need for some mechanism that could help the user on
choosing projects. In the literature, there are other studies on the same context, recommending projects
using different approaches. Although, still there is space for new studies, using new aspects, in an
attempt to verify and validate other results. Therefore, this study focuses on finding relevant projects for
the users based on their interest, on the GitHub, using a set of features with the learning to rank
algorithms support. Analysing the effectiveness of learning to rank, on the recommending projects
context, using the algorithms RankNet, AdaRank and ListNet, in the sample space of 826 repositories and
3464 users on GitHub. The results present the target variable's relevancy and there is still much space for
exploration on learning to rank approach for projects recommendation.