Path-sharing: A new betweenness measure for community identification in networks

Community identification in networks has a wide range of practical applications, including data clustering and social network analysis. We present path-sharing, a new measure of betweenness, for use in identifying densely connected clusters in networks. We show that path-sharing performs well at identifying communities in artificial benchmark networks, giving performance comparable to that of state-of-the-art community identification techniques. We also demonstrate a practical use of path-sharing when used in community identification, by applying it to an image segmentation problem.