EventRank: a framework for ranking time-varying networks

Node-ranking algorithms for (social) networks do not respect the sequence of events from which the network is constructed, but rather measure rank on the aggregation of all data. For data sets that relate to the flow of information (e.g., email), this loss of information can obscure the true relative importances of individuals in the network. We present EventRank, a framework for ranking algorithms that respect event sequences and provide a natural way of tracking changes in ranking over time. We compare the performance of a number of ranking algorithms using a large organizational data set consisting of approximately 1 million emails involving over 600 users, including an evaluation of how the email-based ranking correlates with known organizational hierarchy.