Visual analytics

In the Introduction in chapter 1, we define Visual Analytics (VA) as the science of supporting human reasoning and sense-making via visual representations in which interaction is an essential part of the analysis process. Furthermore, we note that Visual Analytics as a Science is in the borderland between the Behavioral sciences and Computer science. We emphasize that Visual Analytics depends on humans for integrating knowledge, for discovery and insights. The creative processes that are enabled by visualizations are typically not automated but rely on human individuals or teams, trained in both visual literacy and the use of visualization tools, to achieve insights and sense-making. The target audience of this report is FOI-researchers in need of knowledge about visualization. Visualization tools and representations are in many cases deeply engrained in the work processes of the organization, and have evolved organically as people learn to use Visual Analytics tools for performing and communicating their work. New team members need training in understanding the visualizations that are used in the organization. Visual Analytics processes can use, but are not dependent on, the vast collections of data in some organizations and businesses that stereotypically are referred to as "Big Data". Chapter 2 surveys the last three years developments in visual representations and interaction techniques. It demonstrates that age-old wisdom stemming from the most venerable authorities in Visual Analytics is not always based on scientific evidence, and in at least one reported case stands in conflict with recent psychology research. Chapter 3 is about Interactive visualization of multidimensional data. Many applications suffer from the "curse of dimensionality" which means that the number of options to investigate is overwhelmingly large. This is typically due to a so called "combinatorial explosion", caused by each new yes/no choice added in exploring a decision space doubles the number of options to analyze. Chapter 3 reviews two main methods for handling this in visualization: 1) treating all dimensions equally for the purpose of exploration and 2) reducing the number of dimensions by applying mathematical methods such as principal component analysis (PCA) for the purpose of making the most important choices visible among the clutter of options. For each of these main methodological branches several data processing and visualization options are reviewed. Many management situations including business, production, logistics and battle management include a very large number of options. It is often hard to make the full set of options comprehensible to managers in such circumstances. Ideally, Visual Analytics should be employed to guide the manager in the dense forest of action opportunities and possible outcomes. Chapter 4 Effective visualization of multiple options addresses this challenge, finding that the Visual Analytics research community has given it little emphasis. In spite of this, Chapter 4 defines the problem of visually representing multiple options and points to feasible solutions and research directions. Data is captured from interactions with user interfaces, generated by simulations or by measurements. Data is never a precise reflection of reality but is fraught with uncertainty, which could make visual representations of the data misleading. Chapter 5 handles Uncertainty in Visual Analytics focusing on visualization of uncertainty which is the methodology that we need for making the user aware of uncertainties in the underlying data. Chapter 5 reviews appropriate visualization methods for showing the degree of uncertainty in the data, and also reflects on uncertainty in the representations. The level of uncertainty constitutes an extra dimension in the often already multidimensional information, thereby requiring the use of methods for visualization of multidimensional data, as described in Chapter 3. Computational Fluid Dynamics (CFD) is a basic Physical Science underlying much of the design of vehicles, jet engines or ship hulls. CFD is also crucial for defense and security applications, including understanding contaminant flows and underwater applications. Our chapter 6 on Visual Analytics in CFD illustrates how VA is used in applied research. It is pointed out that Visual Analytics is deeply integrated in the work processes, particularly in the postprocessing stage where data from a CFD-simulation is analyzed and compared to experimental data for the purpose of understanding the result. Simulation output is visualized for stimulating human insight. Furthermore, chapter 6 discusses how many organizations have developed in-house visualization tools and representations and mentions the important role of open-source software. The authors' final opinions on Visual Analytics are provided in Chapter 7 Discussion and Conclusions in which we point out that intelligence and creativity in the VA-process is supplied by humans and that 3D Virtual Reality displays will be important for the evolution of VA. Furthermore, we describe applications of VA for defense. In section 7.4 we argue that the reader advantageously can view each chapter in this report as green ectoplasm seeping out from high-dimensional cracks in the shiny walls of the edifice of Visual Analytics.