How to measure an individual influencing others within an online social network in a quantitative way is a critical problem in the field of computational social science. This paper attempts to observe collaborative events occurring at individuals in a social network to obtain such crucial knowledge. We propose a framework with Factorization Machines (FM) to model the social influence among the individuals based on their collaborations; meanwhile, due to the essence of FM, any auxiliary information can be integrated into the modeling process in a straightforward manner. We conduct the experiments on a dataset collected from GitHub, a web-based Git repository hosting service that provides programmers an effective way to collaborate on development projects. In the experiments, we utilize not only the collaborative information among programmers but incorporate various supplementary information, such as user profile (e.g., the number of owned repositories and followers), repository profile (e.g., the number of stars and forks), and textual information (e.g., the title of a repository). The experimental results verify that the effectiveness of the proposed framework on providing better predictive models than several baseline methods. Furthermore, through the experimental results, we observe some interesting social phenomena and provide further analyses and discussions.
關聯:
The 7th Asian Conference on Social Sciences (ACSS’16)，Kobe，2016/06/07~13
如有問題歡迎與系統管理員聯繫 
