Information-assisted data exploration, analysis and visualization techniques

The emergence of web-based scientific simulation portals has enabled scientists to quickly generate large complex scientific simulation data using high performance computing resources. The increasing complexity of these datasets has brought with it challenges of data exploration and analysis. An effective means of exploring scientific volumetric data is through direct volume rendering. Multi-dimensional transfer functions for direct volume rendering have been shown to be an effective means of extracting features and highlighting through the assignment of color and opacity. However, the complexity of setting volume rendering parameters can impede the users\u27 ability to answer relevant scientific questions about their data. This is often due to the fact that designing a proper transfer function does not reflect the scientists\u27 data analysis and exploration process. Furthermore, traditional transfer function widgets provide only limited information about the interaction and correlation of volumetric features since they only present the number of voxels in terms of feature magnitude. This thesis presents an interactive information assisted data exporation and visualization framework for analyzing and exploring scientific simulation data. In this research, we design and develop data exploration and visualization techniques guided by additional information that is processed by analyzing local and global features within input datasets. Features obtained from internal data analysis are utilized to provide initial rendering parameter settings or additional information on top of the conventional user interfaces for data exploration. Our framework provides a semi-automated user interface for transfer function design, modification, and interaction utilizing line charts and contour lines within a slice view for enhancing local data features. We also present a novel abstract attribute space presentation detailing the relationship between the feature space and the volumetric space that leads to better data exploration. In multivariate data exploration, users are presented with a possible sequence for data exploration by employing our dimension ordering scheme and are able to perform logical operations on their selection of feature value ranges while exploring feature space by space. As an extension, our framework includes flow analysis for discrete events such as crime and healthcare reports that allows users to explore events and interactions over space and time to facilitate the discovery of patterns. Finally, we utilize modern GPU computation power and CPU clusters to support interactive data exploration for large datasets