Introduction to Visualizing Hierarchies
Introduction to Visualizing Hierarchies
Hierarchies are a common strategy to structure large amounts of information into nested manageable chunks. We find them everywhere, for example in taxonomies, tables of contents, file directories, or corporate organization charts.
One way to conceptualize hierarchical structures is as inverted trees with the most important, common, or original node at the top, and several other nodes below each node in a recursive way (e.g. organization chart). Alternatives put the root node at the bottom (e.g. genealogical trees), at the left (e.g. directory structure) or the right edge, or in the center with the less important nodes emanating radially. Another way are concepts based on enclosure (e.g. folders containing other folders).
Trees have two properties: the structure of how the nodes are linked, and the content or the attributes describing each node. There are many ways to visually represent trees, depending on whether the emphasis lies on the structure or the content of the tree.
Node-link Diagrams
Node-link diagrams are a very intuitive way to represent hierarchies. Nodes are represented as graphic symbols, and the nodes are connected by lines according to the hierarchical structure. The layout of the nodes can be either cartesian or spherical.
The big disadvantage of node-link diagrams is that they don't scale well with larger number of nodes. Several techniques try to get around this limitation: distortion-oriented approaches (hyperbolic tree), the use of interaction and dynamic pruning, or animated 3D (cone trees).
Node attributes are encoded either as visual properties (size, color, shape) or as icons.
Enclosure-based Diagrams
Enclosure-based diagrams show the most important or root node as a shape that defines the maximum extent of the diagram. Additional child nodes are recursively enclosed by the root node. Enclosure-based diagrams work well to convey the relative importance of the nodes as represented by their size.
Since this representation is a space-filling approach, unlike the very inefficient use of space by the node-link diagram, it scales quite well with the number of nodes.
There are only two parameters to encode the attributes of a node however, color and size, which can be a limitation if there are many node attributes.
Tree maps are prominent examples of enclosure-based representations of hierarchies. They are basically a recursive sub-division of an initial rectangular space. Since they only explicitly show leaf nodes thaty are very effective at using space. The structure of the tree is somewhat difficult to perceive though.
The followig treemap shows 1 million items in one view:
References:
LAMPING, J., RAO, R., AND PIROLLI, P. 1995. A Focus+Context Technique Based on Hyperbolic Geometry for Visualizing Large Hierarchies. In Proceedings of CH '95, ACM Conference on Human Factors in Computing Systems. ACM, New York.
ROBERTSON, G. G., MACKINLAY, J. D., AND CARD, S. K. 1991. Cone Trees: Animated 3D Visualizations of Hierarchical Information. In Proceedings of CHI '91, ACM Conference on Human Factors in Computing Systems. ACM, New York, pp. 189-194.
JOHNSON, B. AND SHNEIDERMAN, B. 1991. Tree-Maps: A Space-Filling Approach to the Visualization of Hierarchical Information Structures. In Proceedings of IEEE Symposium on Information Visualization ‘91, pp. 275-282.
FEKETE, J.-D., PLAISANT, C. 2002. Interactive Information Visualization of a Million Items, Proceedings of IEEE Symposium on Information Visualization 2002 (InfoVis 2002), Boston, USA.