A Comparison of the Readability of Graphs Using Node-Link and Matrix-Based Representations

Node-link diagrams are the traditional standard for graph visualization but suffer from occlusion and link-crossing issues as graphs grow in size and density. Matrix-based representations offer a clutter-free alternative, though they are less intuitive to users.

The study evaluated seven generic tasks on random graphs of varying sizes (20, 50, 100 vertices) and densities (0.2, 0.4, 0.6). The tasks included node/edge counting, finding specific nodes, identifying connections, and path finding. Researchers measured user accuracy and execution time.

Results show that node-link diagrams are effective for small graphs but deteriorate rapidly as size and density increase. Matrix-based representations perform significantly better on tasks like node counting, most-connected node identification, and link finding for larger or denser graphs.

Node-link diagrams remain the default preference for small graphs. However, for larger or dense graphs, matrix-based representations are superior in readability and accuracy. Path-finding remains challenging on both, often requiring interactive enhancements regardless of the chosen representation.