The Haiku system consists of three main components. The first is a machine discovery component (MDC). The system used develops classification rules using a genetic algorithm, but any system making tuple based discoveries, such as classification, characterisation, clustering or association rules could be used. The second, is the discovery - visualisation interface (DVI), which generates a graph from both the data and discoveries made by the machine discovery component. Thirdly, there is HyperSpace [7], a 3D self-organising graph visualisation program.
The discovery process is as follows. Initially, the MDC is used to generate a set of rules which classify the data. Next the DVI generates a graph representing the relationships between the data and the discovered knowledge. This is then passed to HyperSpace which generates the visualisation.