부산대학교 도서관

Social network analysis has been actively pursued to provide an understanding of complex social phenomena. However, graphs used for social network analyses generally contain several errors in their nodes, links, and link weights. In recent years, huge amount of data representing human-to-human interactions are available, and their availability enables us to obtain various types of real social networks. In this paper, we investigate the effect of link weight quantization on the centrality measures in five types of real social networks. Consequently, we show that graphs with high skewness in their degree distribution and/or with high correlation between node degrees and link weights are robust against link weight quantization.