Mining roles of open source software

Most of metrics used in existing OSS evaluation models are difficult to rate and costly to obtain, and the rates of them are tendentiously subjective and inconvincible. This paper tries to find some kinds of data in OSS repositories which can be easily obtained and used as metrics for more practical OSS evaluation. By mining and analyzing nearly 8,000 SourceForge projects, this paper achieves three interesting observations: (1) The Pearson correlation coefficients between the number of roles and the rank of projects are surprisingly high; (2) the centralities of roles reveal that the projects with higher rank have more rational and balanced role structure; (3) the centralities and correlation of roles provide valuable clues for constructing a reference role structure model. These results show that the role structure in OSS projects has a great influence on their quality and provide a new dimension of metrics for more automatic and practical OSS evaluation.