Abstract: The rapid development of open source software ecosystem provides a new model for software development. The research of open source software recommendation system has become an important research field. Most of the existing software engineering recommender systems use the methods of collaborative filtering, machine learning or matching the attribute of the developers between the projects. The research of using network structure and network analysis technology to carry out the recommendation is relatively small. This paper takes GNOME community as the research object, builds a developer-project bipartite network, uses the link prediction technology of bipartite network, recommends the most suitable projects to developers using internal links and compared with the collaborative filtering method. The experimental results show that the method based on the internal links is better than the collaborative filtering method.

Key words: bipartite network, weighted projection, internal link, project recommendation

摘要： 开源软件生态系统的快速发展，为软件开发提供了一种新的模式，对开源软件推荐系统的研究已经成为当前一个重要的研究领域。已有的软件工程推荐系统大都利用协同过滤、机器学习以及开发者-项目属性匹配的方法进行推荐，而利用网络结构和网络分析技术进行推荐的研究相对较少。以软件生态系统GNOME为研究对象，构建开发者-项目二分网络，利用二分网络链路预测技术，采用一种基于内部边的方法对开发者进行项目推荐，并与协同过滤方法进行了对比。实验结果表明基于内部边的推荐方法比协同过滤方法更好。

关键词: 二分网络, 加权投影, 内部边, 项目推荐