Scalable Visual Data Exploration of Large Data Sets via MultiResolution

Daniel A. Keim (University of Konstanz, Germany)

Jörn Schneidewind (University of Konstanz, Germany)

Abstract: During the last decade Visual Exploration and Visual Data Mining techniques have proven to be of high value in exploratory data analysis since they combine human visual perception and recognition capabilities with the enormous storage capacity and the computational power of today's computer systems in order to detect patterns and trends in the data. But the ever increasing mass of information leads to new challenges on visualization techniques and concepts. Due to technological progress in computer power and storage capacity today's scientific and commercial applications are capable of generating, storing and processing massive amounts of data. Most existing visualization metaphors and concepts do not scale well on such large data sets as interaction capabilities and visual representations suffer from the massive number of data points. To bridge this gap, Visual Analytics aim to incorporate more intelligent means than to just retrieve and display the data items to filter the relevant from the non-relevant data. In this context the paper introduces a new approach based on a Multiresolution paradigm to increase the scalability of existing Visual data exploration techniques. The basic idea is to provide relevance driven compact representations of the underlying data set that present the data at different granularities. In the visualization step the available display space is then distributed according to the data granularity, to emphasize relevant information. The paper aims at introducing a technical base of Multiresolution visualization and provides an application example that shows the usefulness of the proposed approach.

Keywords: multiresolution, visual data exploration, visualization technique

Categories: H.0, H.4