Improved Visual Saliency of Graph Clusters with Orderable Node-Link Layouts

Area of Science:

• Graph visualization
• Data analysis
• Information visualization

Background:

• Graphs model relationships, and cluster visualization aids insight discovery in various fields.
• Force-directed layouts enhance cluster visibility but lack intuitive node ordering.
• Matrix layouts offer ordering but lack a node-link metaphor.

Purpose of the Study:

• To investigate the impact of node ordering on cluster visual saliency in orderable node-link diagrams.
• To compare the effectiveness of orderable node-link diagrams against state-of-the-art force-directed graph layout algorithms.

Main Methods:

• Crowdsourced controlled experiment.
• Evaluation of radial diagrams, arc diagrams, and symmetric arc diagrams.
• Comparison with 'Linlog', 'Backbone', and 'sfdp' force-directed layouts.

Main Results:

• Users counted clusters more accurately and faster with orderable node-link diagrams.
• The advantage was more pronounced with low cluster separability and/or compactness.
• Orderable node-link diagrams outperformed tested force-directed algorithms.

Conclusions:

• Node ordering in node-link diagrams enhances cluster visualization.
• Orderable node-link diagrams offer a more effective alternative to force-directed layouts for cluster identification.
• These findings are crucial for applications requiring clear graph cluster representation.