Direct to

Information for

Logo VISUS

Dynamic Graph Visualization

Introduction

Timeline Trees visualization for a market basket data set showing 5 transactions

Visualizing dynamic directed and weighted graphs with an additional hierarchical organization of the graph vertices is a challenging task. Many data dimensions have to be represented at the same time:

  • the graph vertices
  • the adjacency edges induced by the graph
  • the weights of the adjacency edges
  • the inclusion edges induced by the hierarchy
  • the evolution of the graph over time

Traditional approaches use a time-to-time mapping and show the time-varying graph data as animated sequences of node-link diagrams. Though this visualization strategy is very intuitive it also has some drawbacks:

  • if the graphs are very dense, i.e. have many edges, visual clutter occurs caused by many edge crossings
  • animation leads to cognitive efforts for a viewer to preserve his mental map
  • sophisticated layout algorithms are needed to circumwent the two former mentioned problems that have a high run time complexity

Research

In our research we avoid a time-to-time mapping and encode the time dimension into space instead. We use stacked graphical color coded elements to show weighted time-varying relations and we show links only implicitly by different orientations instead of direct explicit links as in node-link diagrams.

 

Layered TimeRadarTrees visualization showing more than 6,000,000 data points of an evolving directed and weighted graph
TimeRadarTrees visualization for soccer match results of 14 years in a part of Europe
The thumbnail view for the goalkeeper showing all weighted relations to all other players in a specific time interval

 

Our approach allows to easily explore a time-varying graph data set for trends, countertrends, and anomalies and has many benefits:

  • visual clutter is reduced by showing the links implicitly
  • cognitive efforts are reduced and the mental map is preserved by using static images
  • interactive features can easily be applied
  • run time complexities are reduced and graphs can be added on-the-fly.

Publications

@inproceedings {p75-burch,
    author = {Burch, Michael and Beck, Fabian and Diehl, Stephan},
    title = {Timeline trees: visualizing sequences of transactions in information hierarchies},
    year = {2008},
    booktitle = {AVI 2008},
    pages = {75--82},
    url = {http://doi.acm.org/10.1145/1385569.1385584},
    doi = {10.1145/1385569.1385584}
}