School of Information Technologies Stream Eb: Stream Edge Bundling

Graph streams have been extensively studied, for instance, for data mining and for stream reasoning, while a fairly limited number of studies have been conducted on visualizations. Recently, edge bundling methods has been extensively investigated to reduce visual clutter in large graph visualizations, however, have focused on static graphs. This paper presents a new framework, namely StreamEB, for edge bundling of graph streams, which integrates temporal, neighborhood, data-driven and spatial compatibility for edges. Amongst these metrics, temporal compatibility and neighborhood compatibility are introduced for the first time, whereas the other metrics have not yet been studied for graph streams. Based on the framework, we present two bundling methods, called FStreamEB (Force-directed Stream Edge Bundling) and TStreamEB (Tree-based Stream Edge Bundling). We demonstrate effectiveness of our framework using US flights data and Thompson-Reuters stock data.