Dynamic networks analysis and visualization through spatiotemporal link segmentation

Abstract

Big network data analysis has become a challenging task not only due to increasing large volume but the appearance of dynamic spatial-temporal relationship. Both network topologies and their link properties are constantly changing as a result of newly established and torn connections. Traditional data mining techniques in large-scale dynamic networks are either incapable or computationally expensive. To that end, we developed a dynamic network analysis and visualization (DNAV) tool. One major component of DNAV is a dynamic graph in which links are divided according to their temporal dimensions. Each segment on network edges represents the dynamic network temporal evolution of graph properties, e.g., locations where the communications occur. Unlike animation approaches, the proposed static view of dynamic networks does not rely deeply on human cognitive ability on remembering changes over different time slots, thus dramatically simplifying the visual analytic process for dynamic networks. To further improve scalability of rendering large networks, data filtering modules such as time selection, hops settings, entities selection, and edge weight thresholds, are adopted in the visualization. Case study demonstrates the effectiveness of DNAV tool in understanding the dynamic network patterns and trends and the potential to analyze anomalies in dynamic communication networks.