Reachability analysis and modeling of dynamic event networks

Abstract

A wealth of graph data, from email and telephone graphs to Twitter networks, falls into the category of dynamic "event" networks. Edges in these networks represent brief events, and their analysis leads to multiple interesting and important topics, such as the prediction of road traffic or modeling of communication flow. In this paper, we analyze a novel new dynamic event graph property, the "Dynamic Reachability Set" (DRS), which characterizes reachability within graphs across time. We discover that DRS histograms of multiple real world dynamic event networks follow novel distribution patterns. From these patterns, we introduce a new generative dynamic graph model, DRS-Gen. DRS-Gen captures the dynamic graph properties of connectivity and reachability, as well as generates time values for its edges. To the best of our knowledge, DRS-Gen is the first such model which produces exact time values on edges, allowing us to understand simultaneity across multiple information flows. © 2012 Springer-Verlag.