Visualizing Dynamic Weighted Digraphs with Partial Links

Abstract

Graphs are traditionally represented as node-link diagrams, but these typically suffer from visual clutter when they become denser, i.e. more vertices and edges are present in the data set. Partial link drawings have been introduced for node-link diagrams aiming at reducing visual clutter caused by link crossings. Although this concept was shown to perform well for some parameter settings, it has not been used for visually encoding dynamic weighted digraphs. In this paper we investigate the problem of visualizing time-varying graphs as one node-link diagram in a specific layout by exploiting the links as timelines. Partially drawn links are used to show the graph dynamics by splitting each link into as many segments as time steps have to be represented. Conventional 2D layout algorithms can be applied while simultaneously showing the evolution over time. Color-coded links represent the changing weights. We use tapered links to reduce possible overlaps at the link target nodes that would occur when using traditional arrow-based directed links. We experiment with different graph layouts and different numbers of data dimensions, i.e. number of vertices, edges, and time steps. We illustrate the usefulness of the technique in a case study investigating dynamic migration data.