Automatic Scatterplot Design Optimization for Clustering Identification
Date created: | Last Updated:
: DOI | ARK
Creating DOI. Please wait...
Description: Scatterplots are among the most powerful and most widely used techniques for visual data exploration of 2D data. Studies have focused on how scatterplot designs can be optimized and suggested guidelines to render effective visualizations. Design choices in visualization, scatterplots in this case, such as the graphical encodings or data aspects, can directly impact the quality of decision making for low-level tasks such as clustering. Compelling visualizations improve understanding of data by leveraging visual perception as well the model based on those understanding. Hence, constructing frameworks that consider both the perceptions of the visual encodings and the task being performed enables optimizing visualization to maximize efficacy. However, we still miss a framework that provides an optimized design framework for effective cluster perception in scatterplots using visual encoding and data aspects. We propose here user-guided automatic tool to optimize the design factors of scatterplot for salient cluster structure. Our interactive tool leverages the application of merge tree data structure to optimize the design decisions on --- sampling rate, sampling algorithms, symbol size, and opacity. We further validate our results with a user study and demonstrate the guidelines that practitioners and designers can extend to other tasks on scatterplots.