Wo ist die Publikation erschienen?Master's Thesis (FH Technikum Wien)
Data visualization has become state of the art when analyzing data in various application domains like production, finance and science, among others. Therefore, research in the field of
data visualization is concerned with the development of novel visualization techniques. Comparatively little research is dedicated to evaluating data visualizations. We looked into existing techniques to evaluate data visualizations on their scalability. Finding, that there is no state of the art method, we propose our own pixel-based evaluation technique. Its goal is to evaluate the scalability of a data visualization with respect to the available screen-space. It works by calculating three different ratios that were based on existing concepts (data-ink ratio, foreground background ratio and the discriminability ratio). The changes of these ratios are then observed over different resolutions to make a statement about the scalability of the visualization. To test our proposed evaluation method, we integrated a v-plot matrix visualization into the software Visplore by VRVis. V-plots are a novel data visualization technique to visualize distribution data and were introduced recently. We used this implementation to export our test data set, that consisted of 375 images containing various differently configured v-plot matrices. The calculation of the proposed ratios was then realized as Python script and applied to the test set images. By observing the resulted trends of the ratios and through conducting hypothesis tests, we found that the ratios indicated that v-plot matrices up to dimensionality 6 could effectively be shown on a resolution of 180x320px and larger. Whereas single v-plots could be displayed effectively on a resolution of 100x160px and above.
Data visualization, Visualization evaluation, V-Plots