UC Santa Cruz

Most graph layout algorithms strive to present an uncluttered view of the graph

that reflects the structural relationship between nodes and edges comprising the

graph. Very few focus on providing a layout based on either node or edge

attribute values. This thesis presents a method that can reflect

structural information, be influenced primarily by attribute values of graph

elements, or some combination of both. This is achieved using a force-based

graph layout strategy and force transfer functions--a flexible graph layout

specification that alters forces depending on attribute values or structural

information. An immediate benefit of this flexibility is the

ability to perform visual clustering via the resulting graph layouts.

As graphs get larger and more complex, the flexibility for exploring different

relational properties of graph elements will allow us to understand them better.

As an example, this technique is used to group left and right blogs as well as detect outliers

in a political blog dataset.