Visual analytics of large weighted directed graphs and two-dimensional time-dependent data

Analysts need to effectively assess large amounts of data. Often, their focus is on two types of data: weighted directed graphs and two-dimensional time dependent data. These types of data are commonly examined in various application areas such as transportation, finance, or biology. The key elements in supporting the analysis are systems that seamlessly integrate interactive visualization techniques and data processing. The systems also need to offer the analyst the possibility to flexibly steer the analytical process. In this thesis, we present new techniques providing such flexible integrated combinations with tight user involvement in the analytical process for the two selected data types. We first develop new techniques for visual analysis of weighted directed graphs. - We enhance the analysis of entity relationships by integration of algorithmic analysis of connections in interactive visualization. - We improve the analysis of graph structure by several ways of motif-based analysis. - We introduce interactive visual clustering of graph connected components for gaining overview of the dataspace. Second, we develop new methods for visual analysis of two-dimensional time dependent data. We thereby combine animation and trajectory-based interactive visualizations with user-driven feature-based data analysis. - We extend guidelines for the use of animation by conducting a perception study of motion direction change. - We introduce interactive monitoring of a new set of data features in order to analyze the data dynamics. - We present visual clustering of trajectories of individual entities using self-organizing maps (SOM) with user control of the clustering process. As a basis for the development of the new approaches, we discuss the methodology of Visual Analytics and its related fields. We thereby extend classification of Information Visualization and Interaction techniques used in Visual Analytics systems. The developed techniques can be used in various application domains such as finance and economics, geography, social science, biology, transportation, or meteorology. In the financial domain, the techniques support analysts in making investment decisions, in assessment of company value, or in analysis of economy structure. We demonstrate our new methods on two real world data sets: shareholder networks and time-varying risk-return data