MLMD: Multi-Layered Visualization for Multi-Dimensional Data

Visualization and data mining techniques have for long been laying emphasis on high-dimensional data processing. In this paper, we propose a multi-layered visualization technique in 3D space called MLMD with its corresponding interaction techniques, for visualizing multi-dimensional data. Layers of point based plots are stacked and connected in a virtual visualization cube for comparison between different dimension settings. Viewed from the side, the plot layers intrinsically form parallel coordinates plots, which are typically effective in visualizing high-dimensional data. MLMD integrates point based plots and parallel coordinates compactly so as to present more information at a time to help data investigation. The user gradually find the desired dimension set of multi-dimensional data by iteratively editing layer dimensions. We carefully design pertinent user interactions for MLMD method to enable convenient manipulation of the layer properties and views. By using MLMD and its mating interaction techniques, proper dimension settings and in-depth data perception can be achieved, presenting a novel way of perceiving multi-dimensional data in 3D visualization space that coordinates multiple views.