Data Sampling in Multi-view and Multi-class Scatterplots via Set Cover Optimization.

Area of Science:

• Data Visualization
• Computer Science
• Statistical Analysis

Background:

• Scatterplots are crucial for visualizing multi-dimensional data.
• Efficient data sampling is essential for handling large datasets in scatterplots.
• Existing methods may not adequately address sampling across multiple views or classes.

Purpose of the Study:

• To develop a data sampling method that optimizes point selection for scatterplots with multiple views or classes.
• To ensure that sampled data (coresets) maintain good approximation guarantees compared to the original dataset.
• To address the challenge of differing data partitions arising from various views or class distributions.

Main Methods:

• Utilized space-filling curves (specifically Z-order curves) to partition point sets.
• Formulated the coreset selection problem as an Exact Cover Problem (ECP).
• Employed an efficient approximate solution for the ECP to select data samples.

Main Results:

• The proposed method effectively partitions point sets using space-filling curves.
• Converting the coreset selection to an ECP allows for efficient processing.
• Quantitative and qualitative evaluations confirm the high quality of the generated samplings.

Conclusions:

• The method provides a robust approach for data sampling in complex scatterplot scenarios.
• Space-filling curves combined with Exact Cover Problem formulation offer an efficient and effective solution.
• This technique enhances the utility of scatterplots for large and multi-faceted datasets.