Methods and Tools for Graph Visualization in Dynamic Systems: Analysis and Experimental Study

Graph visualization is a key tool for interpreting complex data in systems with dynamic relationships, such as transportation, computer, or social networks, where the underlying structure is continuously evolving. In such contexts, there is a growing need for adaptive visualization methods that can ensure clear, intuitive, and timely representation of structural changes. This paper analyzes force-directed graph layout methods, which simulate attractive and repulsive physical forces to determine vertex positions in a two-dimensional space, aiming to minimize the system’s energy function. A custom software application was developed using the Java programming language, integrated with the Spring framework, the GraphStream library, and JavaFX for visualization. The system provides functionality for implementing, configuring, and comparing the Eades, Fruchterman – Reingold, and Kamada – Kawai algorithms. An experimental study was conducted using various types of graphs from open datasets, particularly Rome-Lib and the Scotch Graph Collection, to evaluate the behavior of each algorithm under different conditions. Resultsshow that the Fruchterman – Reingold algorithm demonstrates smooth and gradual layout transitions, high adaptability to structural changes, and good scalability, making it suitable for real-time dynamic graph visualization, such as traffic monitoring systems. The Kamada – Kawai algorithm provides stable and symmetric but has higher computational complexity and less suitability for interactive scenarios due to abrupt movements of individual vertices. The Eades algorithm performs effectively on sparse or tree-like graphs but tends to produce overly long edges and excessive edge crossings in denser graphs. The developed application supports automatic detection of graph structure changes and restarts the layout algorithm accordingly, enabling near real-time reflection of updates in the visual representation. A promising direction for future research is the integration of neural networks to automate layout evaluation, graph-type classification, and algorithm selection based on specific graph characteristics and task requirements. Such an adaptive approach is expected to enhance the efficiency, clarity, and responsiveness of graph visualizations in dynamic systems, contributing to improved monitoring, analysis, and decision-making based on graph- based models.