Significant node identification in social networks

Given a social network, identifying significant nodes from the network is highly desirable in many applications. In different networks formed by diverse kinds of social connections, the definitions of what are significant nodes differ with circumstances. In the literature, most previous works generally focus on expertise finding in specific social networks. In this paper, we aim to propose a general node ranking model that can be adopted to satisfy a variety of service demands. We devise an unsupervised learning method that produces the ranking list of top-k significant nodes. The characteristic of this method is that it can generate different ranking lists when diverse sets of features are considered. To demonstrate the real application of the proposed method, we design the system DblpNET that is an author ranking system based on the co-author network of DBLP computer science bibliography. We discuss further extensions and evaluate DblpNET empirically on the public DBLP dataset. The evaluation results show that the proposed method can effectively apply to real-world applications.