AbstractEuler diagrams are widely used to visualize data items grouped into sets. They can be extended with graphs to visualize networks of data items. This type of data, which we call grouped network data, naturally arises in many areas in the real world. However, existing attempts to automatically lay out combined Euler diagrams and graphs can produce layouts which lead to poor comprehension. Hence, the aim of this thesis is to automatically draw Euler diagrams and graphs in combination that have, compared to the existing state-of-the-art, demonstrably fewer layout properties known to hinder cognition. By achieving this aim, users of visualization techniques for grouped network data will have access to diagrams that are more effective. As a consequence, users will be able to access information that is locked in their data more quickly or more accurately.
To approach this problem we first focus on visualizing sets using Euler diagrams. The Euler diagram well-formedness properties are known to aid comprehension. We devise a new method that ensures the drawn Euler diagrams are well-formed. It is the first method to draw well-formed Euler diagrams from any finite collection of sets, at the expense of, sometimes, introducing non-circular curves, or extra zones. We show practical applicability of our method by bringing it to life in a software tool. We evaluate the diagrams produced by the tool against state of-the-art techniques. In particular, by counting violations of wellformedness properties in other techniques we give an indication of the scale of benefits, with respect to effectiveness, brought about by our new method.
Next, we develop a new method for drawing Euler diagrams and graphs in combination to visualize grouped networks. In addition to Euler diagram properties, much is known about properties of graphs that make them effective. The novelty of our method arises from the fact that is first whose design accounts for effectiveness properties of both Euler diagrams and graphs. A prototype tool is developed and evaluated against state-of-the-art techniques. This comparative evaluation suggests that the diagrams produced by our method display significantly fewer, or no more, violations of effectiveness properties compared to those produced by other techniques, at the cost of increased runtime.
The two new methods are a significant result as there are substantial amounts of set and grouped network data available. Users of visualization techniques now have access to two new techniques which produce set and grouped network visualizations that are, as suggested by our evaluations, more effective for cognition.
|Date of Award||Dec 2019|
|Supervisor||John Howse (Supervisor), Gem Stapleton (Supervisor), Aidan Delaney (Supervisor) & Andrew Blake (Supervisor)|