Interactive exploration of very large relational datasets through 3D dynamic projections

The grand tour, one of the most popular methods for multidimensional data exploration, is based on orthogonally projecting multidimensional data to a sequence of lower dimensional subspaces and then moving continuously from one to another. By running experiments on screen and in the CAVE virtual environment, we were able to use the method for the 3D rendering of very large relational datasets where projections are made onto 3D subspaces. 3D cluster-guided tour is proposed where sequences of projections are determined by centroids of data clusters. It makes inter-clusterdistance-preserving projections in the perspectives that the data clusters are displayed as separate as possible. The exploration of very large datasets is supported by using volume rendering through texture splatting. Volume rendering takes as input only the aggregation or part of data needed for rendering, which can be retrieved efficiently by SQL queries in an ad hoc manner while we explore or drill down large datase...

• Type: Conference
• Year: 2000
• Where: KDD
• Authors: Li Yang
• Keywords: Data Clusters, Data Mining, KDD 2000, Multidimensional Data, Volume Rendering