The data typically found in real-world databases is high dimensional or structured or both, and possesses complex internal relationships. The visualisation of this data, and of discoveries about it is a major problem for the data mining community.
This paper presents the Haiku system, a domain independent visualisation system which is applicable to numeric, nominal and structured data. Haiku also allows the visualisation of many types of discovery.
Haiku forms a 3D visualisations of a graph representing data and discoveries, and the relationships between them. These graphs convey information about the coverage, accuracy and generality of individual discoveries and about the relationships between discoveries. Within data relationships can also be visualised, with similar data forming clusters and outlying data being clearly visible.
We shows an example application visualising the soybean dataset[1] and associated classification rules.