From analysis to interactive exploration: Building 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 suitable 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 interface consists of the navigation structure, a selection of hierarchical visualization techniques, and a set of interaction features. The 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 view adapts itself to display the chosen path and the options valid in the current context. Special effort has been invested in handling non-trivial relationships (e.g., mixed granularity) within hierarchical dimensions in a way transparent to the user. We propose a visual structure called Enhanced Decomposition Tree to to be used 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. Aesthetic hierarchical layout of the node-link tree ensures clear structural separation between the analyzed values embedded in the nodes and their dimensional characteristics which label the links. Our framework provides an intuitive and powerful interface for exploring complex multidimensional data sets. © Springer-Verlag Berlin Heidelberg 2006.