City Research Online

Abstract

This thesis presents a framework and set of readily applicable techniques for enhancing comprehension of complex data visualizations. Central to the work has been the definition and exploration of a new concept, signature exploration.

Visualization is being used increasingly to help make sense of large sets of data and information. Abstractions of complex data can be performed to reduce the dimensions to 2 or 3 for display. Novel or established representations can be used that allow direct mapping of greater numbers of attributes, and of a variety of data structures. There is an ever expanding set of visualization tools available. Two questions face the user: how to choose appropriate displays and how to understand the resultant graphic. This thesis examines how to support the user’s comprehension in this context.

The work makes the following three main contributions to enhancing comprehension of complex data visualizations: the definition and application of signature exploration, a concept describing the exploration of visualization behaviour using specially constructed data; the proposal of a framework for the design of visualization systems for increased comprehension; the introduction of two new forms of interaction - which are here described as visual data tracking and feature fingerprinting.

The central theme for the exploration presented in this work is the notion that a user wants to take data that is known in some way, put this into the visualization process and assess the resultant visual depiction. This intuitive desire has been captured in the definition of the concept, signature exploration. Signature exploration describes the exploration of the behaviour of visual representations using specially constructed datasets that contain features of interest. The datasets are used to explore the signatures of different visual representations and mathematical transformations. The thesis defines and illustrates signature exploration, with five proposed approaches: generic dataset provision; user-construction of data; querying; insertion of landmarks; elicitation and application of feedback data. These applications of signature exploration, together with analysis of the comprehension challenges presented by different aspects of visualization, and established work to support user comprehension, form the basis of the framework for increased user comprehension.

Example software has been developed within the context of a visualization application that employs a number of visualization algorithms to generate graphics for multivariate or proximity data. Principal Components Analysis, Principal Coordinates Analysis and distance metrics of various kinds are the algorithms used. An additional interface is given to the user, to perform signature exploration. The work has resulted in the specification of a set of techniques that developers can readily apply. Two new interaction forms are described: visual data tracking - bi-directional brushing and linking between representations also allowing change of position or value; feature fingerprinting - synthetic additions to real-world datasets to provide the user with calibration of the visual depiction.

Publication Type:	Thesis (Doctoral)
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Departments:	School of Science & Technology > Computer Science School of Science & Technology > School of Science & Technology Doctoral Theses Doctoral Theses

Export

Downloads