[Back]

W. Freiler:
"Set Type Enabled Information Visualization";
Supervisor: H. Hauser, K. Matkovic; Institut für Computergraphik und Algorithmen, TU Wien, 2008; final examination: 2008-03-11.

Information Visualization is a research area in the eld of computer graphics that deals with visual representations of abstract and usually multidimensional data. This data can origin from questionnaires, elections, measurements or simulations. Apart from specialized tools, that are made for a special purpose, there are general purpose tools, that can be used to analyze many kinds of di erent data. These tools are made to handle di erent data types, like numeric or categorical values, some also support more advanced data types, like time series data or hierarchical data. In this document, the data type set will be introduced into the general purpose visualization tool ComVis. A set is a collection of multiple elements, that can also be empty. In many cases, a dimension with the data type set can replace multiple categorical dimensions and make data analysis and exploration more e cient and complex datasets easier to understand. This work will not only explain, how to use sets to explore datasets, but also introduce a new specialized view based on a histogram view, that is dedicated to the use of sets. Of course, most of the already existing views have been modi ed to use sets, otherwise the newly added data type would be di cult to use either. Especially views that can display multiple dimensions were a challenge, because they allow the user to mix sets with other data types. Apart from the use of sets in various views, some additional topics are covered in this document. The conversion of existing categorical data is a very important feature, as well as a fast and e cient data structure. The existing methods for user interaction like brushing and linked coordinated views have to work as expected for all supported data types. A set should not be seen as a new arti cial data type, that we have to convert existing data to, but as the natural data type in many applications. Instead of introducing another conversion step for our data, we can avoid converting data with multiple related attributes to a range of categorical dimensions. Using sets is also an e cient way of dimension reduction, and can reduce the complexity of a dataset, as well as the amount of views needed for exploration. Additionally, there are some examples on how to take advantage of sets when analyzing a real-world dataset. Some special features of this dataset as well as some erroneous entries are easier to nd by using sets and views that support them.