Given an existing visualization e.g. a tree, a map with representations of single data items, how can we secondly group these item according to another different rule? This paper answers this question.
Firstly, the authors surveyed previous methods such as using layout, coloring, clustering, etc. to highlight different group memberships. The method used in this paper is called Bubble Set which resembles grouping items with a bubble that almost exactly contain every single member (other than including non-member items).
One similar technique is convex hull. Similarly, the first step of the two techniques is determining the bubble set boundary. The boundary is not unique, nor minimized. The convex hull results in introducing non-member items while Bubble Set solves this problem by routing virtual edges (edges computed in the first step) to avoid non-member items. The process is a trial and error of moving edge points from within a non-member item to its corner until the point is at a legitimate location.
The technique also offers basic interactions such as adding, moving and removing items. The case study shows several application possibilities such as used on a tree, a timeline, a map and a scatter plot.