Exploring Content-Driven Axis Compression for Visualization of Heterogeneous Data

Area of Science:

• Data Visualization
• Information Visualization
• Computer Science

Background:

• Heterogeneous datasets often lead to visualizations with excessive whitespace and overplotting.
• Sequential data, common in genomics and electronic health records, exhibits density variations, causing occlusion on linear axes.
• Current visualization methods struggle with data density heterogeneity, obscuring crucial data points.

Purpose of the Study:

• To evaluate content-driven axis compression for optimizing data visualizations.
• To minimize whitespace and reduce data point occlusion in heterogeneous datasets.
• To maintain essential distance information while enhancing data visibility.

Main Methods:

• Developed a framework for content-driven axis compression based on data density.
• Compressed axes dynamically according to local data density, requiring minimal prior data knowledge.
• Evaluated the framework through a user study assessing item visibility, accuracy, and data distribution comprehension.

Main Results:

• Content-driven axis compression significantly improves item visibility in visualizations.
• The method maintains high accuracy for tasks related to data distribution understanding.
• Axis compression effectively addresses challenges posed by heterogeneous data density.

Conclusions:

• Content-driven axis compression is a valuable technique for visualizing dense and sparse data.
• The approach enhances the utility of visualizations for complex datasets like those in genomics and healthcare.
• Framework provides guidance on applying axis compression considering data separability and usage context.