SUPPORTING THE UNDERSTANDING OF ASSOCIATION RULE WITH VISUALIZATION
Open Access
- Author:
- Zhao, Hanqing
- Graduate Program:
- Information Sciences and Technology
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- March 28, 2017
- Committee Members:
- Xiaolong (Luke) Zhang, Thesis Advisor/Co-Advisor
- Keywords:
- data mining
visualization
association rule mining - Abstract:
- Integrating advanced data-processing algorithms into visual analytic systems allows analysts to gain more insight into data and to make better decisions. However, such integration also poses new challenges, one of which is the understanding of the algorithms and their results. Sufficient knowledge about what an algorithm is about and how it works could improve the confidence of analysts on analytical results and the quality of final decisions. This thesis reports a study on the use of interactive visualization to support the understanding of the results of association rules, a popular data-mining algorithm. We developed a web-based visual analytics tool that supports the exploration of relations among items, item sets, and rules. We also conducted a qualitative study to examine the factors that affect a user's understanding of data and even the algorithm. We conducted an observational study to let users finish tasks and a semi-structured post-study interview. The findings report that most of the participants found it helpful for them to understand the algorithm better than before, but there are still barriers for. They also shared their thought comparing the visual analytic tool with pure text views. Based on the findings, we make some design recommendations for visual analytical systems targeting for similar issues. For those visual analytic system designers, we suggest that they design system supporting both overview analysis and detailed analysis, and supporting both beginners and deep users with light version and extra add on functions.