Context-adaptive illustrative visualization and analytics techniques

SungYe Kim, Purdue University

Abstract

With advances in information technology, various visualization systems have emerged to help users understand, explore, and analyze their data. Such visualization techniques have been focused on significantly different and wide ranges of contexts (e.g., user, task, data, system, domain, etc.). This indicates that a particular visualization technique would be more effective when applied for a specific task in a specific system environment. Moreover, the adaptation to user and data contexts is essential to provide effective analytics tools. This thesis presents context adaptive visualizations and analytics techniques. In this research, we consider the "context" for analytical tasks using various data types on limited form factors of visualization displays by professional users. We designed and developed visualization techniques to preserve usage contexts, and abstract illustrative representations for various data types and task purposes. We also present a novel representation for the aggregation, abstraction, and stylization of spatiotemporal multivariate data. Two evaluation studies reported that our proposed techniques have significant effects in task performance. Our usage context-preserving techniques help users effectively explore and analyze data sets, and considerably reduce task completion time while maintaining a high accuracy. Our multivariate abstract technique is also significantly better than existing methods in tasks with various levels of complexity. According to the analytical purpose, general tasks are identified. Visualization techniques are developed based on data types and display configurations, and evaluated for effectiveness. Using this approach, we develop categorized visualization solutions to deal with various types of data. The usage context among related document data sets is preserved to help users effectively perform analytical tasks. A multivariate abstract visualization technique is also developed for spatiotemporal data.

Degree

Ph.D.

Advisors

Ebert, Purdue University.

Subject Area

Computer Engineering

COinS