evaluation

Time is a complex dimension, especially when trying to visualize it. In the last ten years a lot of approaches to display and interact with temporal data have been published. They range from linear timeline visualizations to novel ideas employing visual metaphors and even clustering techniques to support the user in exploring large-scale data sets. The diversity of the proposed methods has raised the awareness that a common categorization needs to be defined to efficiently evaluate the usability and interactivity of information visualization tools.

Therefore, this work aims at giving a detailed overview of the possibilities and problems of current information visualization tools by applying a recently published categorization. A data set containing air pollution data measured in the years 2002 to 2006 at five different measurement stations in Great Britain is displayed with each of them. To enhance the judgement of the visualization tools, tasks that cover different areas of practical work are defined and carried out. After this practical part the categorization is applied to all of the examined applications. Both the task accomplishment and the use of the categorization are then reflected and occurred problems are described. Possible improvements are pointed out and future research areas are mentioned.

This master thesis describes the evaluation of an interactive information visualization technique that is capable of displaying quantitative attributes (numeric values) of multivariate data over time and corresponding qualitative abstractions (interpretations) of the quantitative values (SemTimeZoom). The integration of interpretations and a-priori knowledge in the form of qualitative abstractions is especially useful in the medical domain. Vital parameters of patients can be analyzed using predefined domain knowledge and the resulting interpretations can be visualized together with raw numerical measurements.

The investigated visualization technique uses different visual representations of the data depending on the vertical display space of a single parameter and combines the quantitative and qualitative attributes of a parameter into one combined representation. The area-aware method to display different representations is called semantic zooming. Although the developed visualization technique appears very promising, it has not yet been evaluated. Novel visualization techniques need to present measurable benefits to encourage more widespread adoption. To assess the effectiveness of this visualization technique, a comparative study was performed. The visualization technique that was used for the comparison is also capable of displaying raw quantitative values and qualitative abstractions but uses static and separate visual representations for quantitative and qualitative attributes of the data. The comparative study was conducted by means of a controlled experiment that revealed faster completion times especially for more complex tasks involving comparison of quantitative values within specified qualitative categories in favor of the SemTimeZoom technique. All tasks that were used in the experiment involved the qualitative attributes of the data to evaluate the effectiveness for exploratory data analysis with qualitative abstractions. It is generally acknowledged in the information visualization research field that it is necessary to evaluate visualization techniques, but the difficulties of conducting such evaluations still remain an issue. In the course of this study, evaluation functionality was integrated into the Java software prototypes that were used for the controlled experiment. A software library was built based on the evaluation functionality to facilitate future evaluation studies. Care has been taken to develop an easy-to-use, flexible and reusable software library that can be integrated into other prototypes that need to be evaluated.