Stat-Check: A Responsible, Data-Driven Visualization Recommendation Systems

Multiple state-of-the-art visualization tools, such as D3 and Vega- Lite, emerge to make visualization generalization easy for researchers, data scientists and even for users without programming experience and background knowledge. People could largely benefit from these easy-to-use systems and tools to create visualizations and run analyses. However, there are data issue detection and data issue cleaning either not supported in the tools or can not be easily used. Moreover, there is a need that visualization systems should not only to output required charts but also to flag and explain potential issues within users’ data. Therefore, based upon a previous prototype, we present a visualization recommendation system which can detect potential statistical data issues and suggest possible solutions. We propose refined methods on building collection of implementable and commonly-seen data issues, a function that accepts datasets from the user’s side, an interactive user interface that is in a clear layout, and an educative visual recommendation panel including necessary explanations and possible solutions for detected data issues. We also present a user study involving 4 participants to validate the effectiveness and usability of our system.

Chen Chen
Chen Chen
3rd-year graduate student

My research interests include information visualization, human-data interaction, and eXplainable AI (XAI).