Using epidemic betweenness to measure the influence of users in complex networks