KAVA-Time

Analytical reasoning for real world problem solving involves large volumes of uncertain, complex, and often conflicting data that analysts need to make sense of. In this context, time-oriented data is commonplace and plays a special role. Due to the distinct characteristics of time, appropriate methods for exploration and analysis are needed. Visual Analytics provides sophisticated methods that combine interactive visual interfaces with automated analysis methods. Ideally, a Visual Analytics environment would adapt itself to the user`s context and domain specifics of the data to analyze. For example when displaying lab results in electronic patient records this could mean to show expected value ranges for healthy patients depending on their context such as gender or age. For representing stock price data, the time axis would suppress weekends and bank holidays to avoid a distorted representation of value change. In a scenario of exploring energy usage in a building, a representation would be chosen that accounts for daily and weekly cycles by aggregating them and aiding to detect the unknown. Solutions like these might be achieved by creating specialized applications for each domain and analysis problem at hand. However, this would cause lots of effort and make maintenance and reuse difficult. To avoid this, we will design Visual Analytics methods that accommodate to different contexts. Throughout this project, we will study how we can take advantage of explicit expert knowledge in the Visual Analytics process to make analytical reasoning more effective and efficient. We plan to develop and evaluate knowledge specification methods as well as knowledge-assisted visualization and interaction methods for time- oriented data. This encompasses two main objectives: (1) to capture analysts` domain knowledge and explorative interests, and (2) to take advantage of the explicit knowledge in interaction and visualization methods. Current approaches mostly rely on static, externally given knowledge and do not emphasize reuse and sharing of these specifications. In contrast to that, we aim to integrate specification methods directly with interactive Visual Analytics methods to allow intuitive and direct refinement of explicit knowledge by analysts. The visualization methods will make use of explicit knowledge by automatically adapting themselves and using abstractions of the input data. For development and evaluation we plan to adopt data and user tasks from an application scenario in medicine. Tackling this issue will give rise to more effective environments for gaining insights - the possibility to specify, model, and make use of auxiliary information about data and domain specifics in addition to the raw data, will help to better select, tailor, and adjust appropriate methods for visual representation, interaction, and automated analysis.

Seeing and understanding gaining insight through visual analytics. Analyzing and interpreting large amounts of data is often difficult. However, by using the appropriate analytical approach one may discover information that would have remained well hidden otherwise. Crucial for discovering insights is the interplay between automated data analysis through a computer and the interpretation of the data by experts using interactive visualization tools. Knowledge-Assisted Visual Analytics for Time-Oriented Data (KAVA- Time), a project funded by the FWF, was aimed at developing appropriate software tools for data analysis. Interplay between humans and computers. Whether it is data about a patients health status, statistics on climate change or behavioral patterns of malware, in many instances experts and scientists are forced to analyze large amounts of data. Only through detailed analysis and interpretation of data, valuable findings can be gleaned. However, such an analysis is not always easy. Often there is an unmanageable amount of data or even contradictory results. Using computers, it is possible to discover patterns and trends, and visualize data. Computer software often fails to go beyond trivial patterns due to a lack of background knowledge. At this point human experts are needed. Only experts are able to interpret data and place them in an appropriate context. In the KAVA-Time project, concepts have been developed that foster a knowledge-assisted visualization of time-oriented data. In doing so, human expert background knowledge can be incorporated in order to improve the visualization and processing of data. Consequently, experts are enabled to interactively capture their knowledge during the process of data analysis and incorporate it in a computer system. Based on a newly developed theoretical model for knowledge-assisted visualization and analysis it was possible to create two exemplary use-cases for a concrete implementation. Discovering malware through behavioral patterns. Large-scalecyber-attackson critical IT infrastructure are a common occurrence these days and computer security is threatened through newly emerging malware and viruses. In the project KAVA-Time the project team has developed methods to support the analysis of malware based on characteristic behavioral patterns, by creating a prototype and evaluating its applicability. In doing so, it became evident that experts benefit in a number of ways from explicit knowledge. Analyzing and visualizing clinical data in walking patterns. Many people are affected by dysfunctions in their walking patterns (gait abnormality) due to functional deficits. In order to adequately support the therapists in their work a knowledge-assisted software tool KAVAGait was developed and prototypically tested. The tool facilitates the diagnosis of a gait abnormality using complex measurements.