Incorporating geographical location for team formation in social coding sites

With the proliferation of open source software and community, more and more developers from different background (e.g., culture, language, location, skill) prefer to work collaboratively and release their works in social coding sites (e.g., Github). Given a collaborative project with a set of required skills, it is an important and challenging task to form a team of developers that have not only the required skills but also the minimal communication cost. Previous works mainly leverage historical collaboration records among team members to model the communication cost, while ignoring the impact of geographical location of each developer. In this paper, we aim to exploit and incorporate the geographical information to improve the performance of team formation in social coding sites. Specifically, we conduct two objective functions for the collaboration records and geographical proximity correspondingly, and propose two optimization algorithms. Comprehensive experiments on a real-world dataset (e.g., GitHub) demonstrate the performance of the proposed model with the comparison of some state-of-the-art ones.