Predictive Team Formation Analysis via Feature Representation Learning on Social Networks

摘要: Team formation is to find a group of experts covering required skills and well collaborating together. Existing studies suffer from two defects: cannot afford flexible designation team members do not consider whether the formed truly adopted in practice. In this paper, we propose Predictive Formation (PTF) problem. PTF provides flexibility designated delivers prediction-based formulation compose team. We methods by learning feature representations based on node2vec [4]. One Biased-n2v that models topic bias each expert social network. The other Guided-n2v refines transition probabilities between guide random walk heterogeneous graph expert-expert, expert-skill, skill-skill. Experiments conducted DBLP IMDb datasets exhibit our can significantly outperform state-of-the-art optimization-based approaches terms prediction recall. also reveal with tight connections lead better performance.