The Classification Structure of Methods for Large-Scale Graph Visualization

Abstract

The article addresses the current challenge of large-scale graph visualization, which is accompanied by issues such as difficult visual perception due to the high number of vertices and edges, the complexity of interpreting visual representations, and the low efficiency of existing methods. The study analyzes existing visualization methods and concludes that there is no strict hierarchical classification, which complicates their systematization, selection, and identification of promising research directions. The goal of this research is to develop the classification structure for existing large-scale graph visualization methods. Based on an analysis of recent studies and publications, three main categories of methods solving key visualization tasks are identified: vertex and edge coordinate assignment methods (including force-directed, spectral, genetic, and graph neural network-based methods), large graph processing methods (covering graph modification, subgraph sampling, hierarchy creation, parallel computing, and machine learning methods), and graph visual representation optimization methods (including edge crossing minimization, vertex distance maximization, and adaptive cooling methods). The proposed classification structure details each of these categories into subcategories, considering differences in approaches to solving various aspects of large-scale graph visualization. The developed structure systematizes existing knowledge and provides a foundation for further comprehensive method classification, facilitating informed method selection based on the input graph and identifying directions for developing new, more efficient large-scale graph visualization methods