Quality-aware visual data analysis

The quality, certainty, or confidence of decisions made during the visual analytics process depends on many factors, including the completeness and reliability of the initial data, information loss due to filtering, sampling, and other transformations, and the accuracy and clarity of the visual presentation. Unfortunately, in most visualization tools the analyst is unaware of these and other forces that degrade the meaningfulness of their results. In this paper, we describe our efforts to design strategies for tackling the measurement, display, and utilization of quality aspects at all stages of the visualization pipeline. The goal is to help analysts maintain an awareness of the accuracy and completeness of the information conveyed in the images, and subsequently the patterns observed and decisions made based on the analysis. Quality measures can be used both to assist analysts in selecting, transforming, and mapping their data as well as to automatically refine processes to generate higher quality views. We have implemented several such techniques within XmdvTool, a public-domain package for visual analytics. We illustrate the quality-specific components of our tool with several case studies to show the usefulness of the approach. We also describe user studies that were performed to validate the accuracy of our quality measures. This work is supported under NSF grants IIS-0119276 and IIS-0414380. M. Ward (B) · Z. Xie · D. Yang · E. Rundensteiner Computer Science Department, Worcester Polytechnic Institute, Worcester, MA, USA e-mail: [email protected] Z. Xie e-mail: [email protected] D. Yang e-mail: [email protected] E. Rundensteiner e-mail: [email protected]