Do more people make the code more defect prone?: social network analysis in OSS projects

Abstract

A software quality model consists of three pillars: 3Ps (Product, Process, and People). Many software quality models in the literature has so far taken process and product related attributes into consideration, but there are very few studies that model people. In Open Source Software (OSS) development projects developers and their interactions with each other and the software product are even more important than closed loop projects. In open source software loosely defined processes usually increase the significance of people. In this research we are motivated to explore the social network of people in OSS projects. We extracted the social network metrics of developers in four large-scale open source software to examine their relation to software quality in twofold. First, we analyzed the effect of social network metrics in predicting post release defects. Second, we explored the effect of social networks to defect proneness of open source software. Our experiment results showed that social network metrics do not have a distinct impact over and above static code and churn metrics in predicting defects. On the other hand, our results revealed that there is a positive correlation between social network complexity and defect proneness.