University of Konstanz
Graduiertenkolleg / PhD Program
Computer and Information Science

Publications

Back to Publications

Author(s) Vinnik, S., Mansmann, F.
Title From analysis to exploration: Building enhanced visual hierarchies from OLAP cubes
Abstract We present a novel framework for comprehensive exploration of OLAP data by means of user-defined dynamic hierarchical visualizations. The multidimensional data model behind the OLAP architecture is particularly well suited for sophisticated analysis of large data volumes. However, the ultimate benefit of applying OLAP technology depends on the "intelligence" and usability of visual tools available to end-users. The explorative framework of our proposed user interface consists of the navigation structure, a selection of hierarchical visualization techniques, and a set of interaction features. The novel navigation interface allows users to pursue arbitrary disaggregation paths within single data cubes and, more importantly, across multiple cubes. In the course of interaction, the navigation structure adapts itself to reflect the chosen path and the options valid in the current context. Special effort has been invested in handling non-trivial relationships within hierarchical dimensions in a way transparent to the user.
We propose a visual structure called Enhanced Decomposition Tree to use along with popular "state-of-the-art" hierarchical visualization techniques. Each level of the tree is produced by a disaggregation step, whereas the nodes display the specified subset of measures, either as plain numbers or as an embedded chart. The proposed technique enables a stepwise descent towards the desired level of detail while preserving the history of the interaction. Additionally, aesthetic hierarchical layout of the node-link tree ensures clear structural separation since the analyzed values are embedded in the nodes whereas their dimensional characteristics label the links. Our framework provides an intuitive and powerful interface for exploring complex multidimensional data sets.
Download ViMa06.pdf

Back to Publications