Graham J. Wills
Visualization Research
Bell Laboratories (Lucent Technologies)
USA
gwills@research.bell-labs.com
The difference between displaying networks with 10-1000 nodes and displaying ones with 10,000-100,000 nodes is not merely quantitative, it is qualitative. Layout algorithms suitable for the former are too slow for the latter, requiring new algorithms or modified (often relaxed) versions of existing algorithms. The density of nodes and links displayed per inch of screen real estate requires special visual techniques to filter the graphs and focus attention. Compounding the problem is that large real-life networks are very often weighted graphs and almost always have much additional data associated with the nodes and links. A system for investigating and understanding such large, complex data needs to be able to display both graph structure and node and link attributes so that patterns and information hidden in the data can be seen. In this talk we describe a tool that addresses these needs, the NicheWorks tool. We describe and comment on the available layout algorithms, the linked views interaction system and detail examples of the use of NicheWorks for analyzing web sites and in detecting international telephone fraud.
NicheWorks is a visualization tool for the investigation of very large graphs. It allows the user to examine a variety of types of attributes of the graph nodes and links in conjunction with their connectivity information. By very large we mean graphs for which we cannot look at the complete set of labeled nodes and links on one static display. Typical analyses performed using NicheWorks have had between 20,000 and 1,000,000 nodes. On current mid-range workstations a network of around 50,000 nodes and links can be visualized and manipulated in real time with ease. Categorical, textual and continuous attributes can be explored with a variety of one-way, two-way and multi-dimensional views as can their relationships to the graph structure.