UC Davis

High-dimensional data analysis is a major target application for visualization. When analyzing high-dimensional data, dimensionality reduction (DR) plays a pivotal role as it uncovers the intrinsic features of the data. Current DR methods, however, provide little support in terms of the interpretability in their results, usability with respect to interactive visualizations, and flexibility for handling various data types. This dissertation investigates the problem stated above and presents a set of novel DR methods coupled with interactive visualization, in which four different topics are addressed: (1) interpretation of DR results, (2) analysis of time-dependent multivariate data, (3) design of multivariate streaming data visualization, and (4) introduction of a new approach to comparative network analysis. Illustrative case studies demonstrate the new capabilities greatly enhance the collective power of visual analytics and DR.