Optimizing occlusion appearances in 3D association rules visualization

Abstract

Providing efficient and easy-to-use graphical tools to users is a promising challenge for data mining (DM). Visual data mining (VDM) is a new and active research area which goal is to provide powerful and suitable tools for data miners. Some graphical tools have been developed to extract and visualize association rules (AR), among which a three dimension representation where the x-axis is the AR premise, the y-axis is the AR conclusion and the z-axis is a metric value of AR. The 3D approach is one standard representation that is often implemented in many DM tools. However this approach suffers from an overlapping between several objects in the 3D space making some objects unseen or partially truncated. This problem is known as the occlusion problem. In this paper, we propose to formalize it as an optimisation problem of occlusions. Then we define conditions to limit occlusions and finally we propose different heuristics based on ordering of axis-elements, to considerably reduce the number of generated occlusions.