| The primary challenge of computing society in the current and coming decades is not computational power; it is dealing with the collected or generated data in the computers, especially in the incomprehensible form in which they are currently presented to users. In most management domains, problem-solving is overwhelming because of the large volume of complicated data, multiple complex relationships among data, and the negotiability of the constraints. Efforts have been made in introducing graphical representations. However, little research has resulted in the development of effective visual representations for large-volume, multi-dimensional, non-geometric-based data for human beings to gain insight into the data. It is extremely vital and urgent to improve the communication between users and computers, to transform the vast computing-related data into comprehensible representations that help humans to understand the data. Humans are visual creatures. Most of what we learn comes through our sight. An effective solution to the challenge is to shift some of the user's cognitive load to the human visual perceptual system by using computer-generated, domain-specific visualizations.; This dissertation focuses on developing a research strategy for building visualizations of non-geometric data that are massive in both volume and dimensionality, to help decision-makers to achieve data comprehension and eventually to improve problem-solving performance. The research first took real problems from a specific business domain--the manufacturing production planning domain--to explore the feasibility and special concerns of visualizations for decision-making support. A prototype system VIZ{dollar}sb-{dollar}planner was developed under X-window environment. It was based on human problem-solving processes, captured the multi-dimensional nature of the problems in easy-to-understand visual representations, and was intended to improve the performance of planners' entire problem-solving process. A lab experimental study showed that users using a visualization system generated more potential solutions and had more efficient changes in raw data than those using a traditional computer support system did. They also were more confident about the tasks they were going to do and more satisfied with the outcomes than those using the traditional system were. A general model for building domain-specific visualizations for management domains is then proposed. As the visualization of large data sets is a problem of concern in many management domains, the dissertation research makes both theoretical and practical contributions. |
