中文摘要：

根据开发者的历史开发信息，推荐潜在合作对象或能够胜任的任务，对于减少开发延时，提高团队开发效率有极大的促进作用．大量已有工作主要研究单个开源项目开发者社区，通过确定开发者的个人技能实施推荐．然而，以整个开源社区为视角，研究使用开发者实践技能做推荐，实现开发者跨项目与组织边界合作的工作相对甚少．为丰富这一工作，文中结合开发者的实践技能相似性与共同开发者数，为Sourceforge．net社区未曾合作的开发者提供一种同行推荐的方法．研究结果表明文中的方法能够有效地实现开发者同行推荐，相比只考虑共同开发者数时成功推荐的开发者数增加了31．1％，且推荐精度提高0．4倍，缓解了社区合作的稀疏性．最后，应用文中提出的方法开发了一个Roster推荐系统，为社区开发者的交流建立了一种新的辅助途径．

英文摘要：

Applying developer’s historical activities to recommend the potential collaborators orcompetent tasks plays an important role in software management which can greatly improve thedevelopment efficiency and further reduce the unnecessary delay.Though there are numerousliteratures on this topic,they only focus on recommendation in the context of a single open-sourceproject.Few explored the recommendation problem that is to promote the cross collaborationbetween developers engaged in different projects and organizations from the perspective of the wholeopen-source community.In this paper we bridge this gap,by proposing an approach to recommendpeers for developers who do not collaborated with each other now but will be most likely to constructcooperation in the future according to their similar expertise and the number of common developers.Empirical results on the Sourceforge.net dataset show that our approach can achieve an effective peerrecommendation for developers.Compared with those only considering the number of commondevelopers,in our approach,the number of successful recommendations is increased by 31.1%and the accuracy improved by 40%.Finally,our approach has been automated by a tool Roster.