Högskolan i Skövde

Large quantities of data are being collected and analyzed by companies and institutions, with the intention of drawing knowledge and value. Advances in storage, computation, automated analysis and visual and interactive techniques have facilitated this process. It is, however, not always transparent on how these can be brought together for the effective and efficient exploration, monitoring and/or processing of data. That is, knowing which automated techniques and frameworks to use for a given task, how to deploy them and integrate them, how to interpret their results and processes, and how to ease their use through visual and interactive techniques.

In the context of exploratory data analysis, where users approach data without preconceived hypotheses, this thesis proposal argues for the benefit of a framework describing the components, techniques and relations that contribute to the visualization and explanation of cluster patterns in large and multidimensional datasets. That is, a Visual Analytics framework aimed at supporting data scientists in their first steps towards understanding the overall structure of a dataset. The problem area is large and, therefore, the scope is limited to the visualization of cluster patterns in table-like datasets with meaningful attributes.

This thesis proposal motivates the relevance of conducting research for the development of such framework. It formalizes a set of research questions and presents a research plan based on the design science research method. Moreover, it provides a description of preliminary results as well as related background theory and state-of-the-art research.