A Tie Strength Model For Reconstructing Collaboration Networks on GitHub A study of the Ruby on Rails project network

Abstract

Social network analysis based on follow relations, on GitHub and other social networking websites, ignores the strength and time of relations. This paper introduces an approach to reconstruct weighted social networks covering extensive periods of time using communication data. A tie strength model is suggested to determine edge weights that reflect the intensity of collaborations. The approach and the model are tested by investigating the 10 year history of Ruby on Rails, an open source software project. The results show that for a group of 283 software developers, social network analysis based on communication data is more successful at finding collaborators compared to analysis based on follow relations. The model for tie strength results in a significantly better ranking of the intensity of collaborations compared to rankings based on sums of activity. The approach in this paper allows GitHub activities to be combined into meaningful project networks that can be used to answer more fundamental questions about open source software development, e.g. determining if some or any aspect of OSS communities and development are constant and predictable.