Integrating Explicit Knowledge in the Visual Analytics Process: Model and Case Studies on Time-oriented Data

Abstract 

Visual analytics (VA) aims to combine the strengths of the human user and computers for effective data analysis. In this endeavor, the user’s implicit knowledge from prior experience is an important asset that can be leveraged by both, the user and the computer to improve the analytics process. While VA environments are starting to include features to formalize, store and utilize such knowledge, the mechanisms and degree to which these environments integrate explicit knowledge varies widely. Additionally, a theoretical model and formalization of this class of VA environments is not available in the VA community yet. This doctoral thesis aims to close this gap by proposing a new theoretical high-level model conceptually grounded on the ‘Simple Visualization Model’ by Van Wijk supporting the visualization community. The new ‘Knowledge-assisted VA Model’ provides the ability to describe all components and processes to characterize knowledge-assisted VA systems. Additionally, it supports visualization experts and designers by comparing and evaluating knowledge-assisted VA systems as well by creating new solutions. To demonstrate the model’s application, we use problem-driven research to study knowledge-assisted visualization systems for time-oriented data in the context of two real world problems. The first case study focuses on the domain of IT-security to support experts during behavior-based malware analysis. Therefore, we developed KAMAS, a knowledge-assisted visualization system for behavior-based malware analysis, describing its design, implementation, and evaluation. Additionally, to support clinical gait analysts during their daily work, we conducted a second case study developing KAVAGait, a knowledge-assisted VA solution for clinical gait analysis. In addition to applying the ‘Knowledge-assisted VA Model’ in two case studies, we also elaborate on two examples from literature. Moreover, we illustrated the utilization of the model for the comparison of different design alternatives and to evaluate existing approaches with respect to their use of knowledge. Our model provides the opportunity to inspire designers by using the model as a high-level blueprint to generate new VA environments using explicit knowledge effectively. Additionally, we observed that the VA process benefits in several ways by explicit knowledge: 1) by including it into the automated data analysis process; 2) for adapting the system’s specification and 3) to faster gain new implicit knowledge about the data. Finally, we present possible future directions for future research on the integration of explicit knowledge in VA.

Download 

20170623_Dissertation_Markus_WAGNER_compressed.pdf