GraphOptima: A graph layout optimization framework for visualizing large networks

Abstract

Graphs illustrating online networks, participants, and interactions have become essential tools for studying issues like misinformation, botnets, algorithmic filtering and many other areas of research. Researchers use these visual representations to examine underlying structures, form hypotheses, and share their findings. However, achieving visually appealing network visualizations often involves manually testing several layout algorithms and fine-tuning their parameters. Furthermore, due to the computational complexity of rendering large networks, there is also usually a long wait time between parameter tests. This paper introduces GraphOptima, a framework for optimizing graph layout and readability metrics. GraphOptima automates parameter selection, layout computation, and readability metric calculation. Rather than providing a single ‘optimal’ solution, the framework generates a range of solutions under different parameters, enabling researchers to explore multiple solutions based on multi-objective optimization. The framework supports parallel layout calculations without modifying the layout algorithm, efficiently managing computational resources in high-performance computing environments. In addition to introducing a new framework, the paper shows how GraphOptima can be used to explore distinct layouts for three sample networks and to decide on trade-offs among three readability metrics: crosslessness, normalized edge length variance, and min angle.