Abstract

Efficient exploration of large and complex datasets, as for instance, time-series and provenance data, is an ongoing research challenge in visual analytics. Visualizing such datasets in one go often leads to visual clutter, making it hard for users to identify potentially interesting data subsets. A possible solution to reduce the clutter is Focus+Context techniques, which visualize selected regions in greater detail while preserving an overview with reduced details. For large datasets, however, selecting focus regions can become a timeconsuming task if each region must be selected individually. Furthermore, in the case of temporal data, the interest in a particular data subset might not remain constant but, rather, shift over time or switch to other data subsets. Consequently, it is necessary to develop Focus+Context solutions tailored to large temporal data. This thesis presents four interactive visualization approaches for highlighting potentially interesting subsets in time-series and provenance data. The solutions utilize modular degree of interest functions that are driven by one or multiple data attributes, the topology of the graph, or a combination of both. The practical applicability of these approaches is demonstrated by means of different case studies from cloud computing, finance, and biomedical research.

Citation