The GEMS 2.0 project is a DFG-funded research project that ran from July 2017 to December 2020. It was a direct follow-up of the first GEMS project, which was also conducted by the Chair of Visual Analytics of University of Rostock and the Interactive Media Lab Dresden of Technische Universität Dresden.
The project is motivated by the fact that today’s world is a data-driven one. As a result, decision-making processes are based on the analysis of data and communication of derived insights. Such data can be arbitrarily complex, with data elements being characterized through many attribute dimensions as well as standing in specific relations to each other. The notion of multivariate graphs address such complex data that features both relational and multivariate aspects. Within the GEMS 2.0 project, the Chair of Visual Analytics and the Interactive Media Lab Dresden aim to better investigate this area.
Specifically, we explored how novel visualization and interaction concepts can ease working with such data sets, particularly with respect to editing, comparison, and usage on interactive displays. Within this joint project by the University of Rostock and the Technische Universität Dresden, we brought together expertise on visualization and interaction research, resulting in multiple novel approaches for working with multivariate graphs. Our explorations are part of the broader research area of interactive visual data analysis, emphasizing that suitable interaction and visualization means are crucial for working in data-driven processes.
Matrix Visualization for Multivariate Graphs
Within the project, we found that matrix visualizations are a promising approach for multivariate graphs. While such matrices have been used in the form of adjacency matrices to indicate relations between entities, they have been rarely considered to show the multivariate aspects. By splitting the matrix in two triangular halves, we presented a new way to show both relational and multivariate aspects at the same time. Specifically, the multivariate half shows a similarity measure calculated over the attributes between node pairs.
Such a matrix is particular suited to provide an overview on the graph. As a novel interactive technique called Responsive Matrix Cells, we developed a focus+context approach building upon such a matrix visualization. Here, embedded visualizations are provided for a selection of cells and can seamlessly be zoomed to access more details. An important aspect of this approach is the possibility to offer integrated editing facilities, allowing for directly manipulating the embedded visualizations; among others, this enables analysts to fix values or run what-if analysis.
An online version of our Responsive Matrix Cells system is available at https://vcg.informatik.uni-rostock.de/~ct/software/RMC/.
Visual Data Analysis across Devices
Visual data analysis - including graph analysis - is not only taking place on traditional devices anymore. Therefore, we considered a wide range of modern devices reaching from mobile ones to large displays as well as their combination. To facilitate the analysis on all devices, we focused on the notion of responsive visualization, i.e., visualizations views that can adapt to different device contexts. We started with general considerations of important aspects and, then, integrated aspects of it in our matrix approach.
In addition, we also investigated how data analysis can be conducted within multi-device environments, i.e., with multiple devices in combination. Specifically, we developed new interaction techniques, e.g., how to interact from varying distance with secondary devices, as well as approaches for automatically distributing components of a visualization interface across the used devices.