Résumé
Experienced developers are key for the success of software development projects. In open-source software development, due to openness and distance, one cannot always rely on interpersonal interactions to know who these key people are. Automating the mining of experienced developers is not an easy task either, because of the subjectivity and relativity of what experience is and also because the material to search from (code and development-related metadata) does not obviously relate developers to their capabilities. Some research works propose developer profiling or clustering solutions though, from which we take inspiration. This paper advocates that it is possible to learn from tangible metrics extracted from code and development-related artifacts who are the experienced developers. It uses a supervised learning-based approach trained with a manually labeled dataset of 703 developers from 17 open-source projects from GitHub for which 23 metrics are automatically extracted. Experienced developers classification results show a high F1 measure. A companion explainability study analyzes which metrics are the most influential.