Graph and Network Visualization

Graph and network visualization depicts entities and their relationships, most commonly as node-link diagrams, helping people see structure such as clusters, paths, and central nodes.

Definition

Graph and network visualization is the visual representation of data consisting of entities and the relationships among them, using layouts that position nodes and route links, or matrix representations, so that structural properties of the network become perceptible.

Scope

This topic covers the visual representation of relational data: node-link diagrams and graph layout algorithms such as force-directed and hierarchical layouts, matrix representations of graphs, techniques for large and dense networks, and navigation and aesthetics criteria for readable graphs. It does not cover the algorithmic theory of graphs as data structures, which belongs to algorithms, nor the perceptual basis of encoding in general, treated under visual encoding and perception.

Core questions

• How do layout algorithms position nodes to reveal structure?
• When are node-link diagrams better than matrix representations?
• How can very large or dense networks be made readable?
• What aesthetic criteria make a graph drawing easy to interpret?

Key concepts

• node-link diagram
• force-directed layout
• hierarchical (layered) layout
• adjacency matrix representation
• edge crossing minimization
• graph aesthetics
• clustering and aggregation
• scalability of large networks

Key theories

Graph layout and aesthetics

Graph drawing seeks layouts that satisfy aesthetic criteria such as few edge crossings, short uniform edges, and revealed symmetry; force-directed and hierarchical algorithms produce layouts that make structure visible.

Node-link versus matrix representations

Controlled experiments show node-link diagrams are good for small, sparse graphs and path-following tasks, while matrix representations scale better to large dense graphs and tasks like finding the most connected node.

Navigation and scalability

Surveys of graph visualization catalogue methods for navigating and simplifying large networks, including clustering, filtering, and focus-plus-context navigation, to keep big graphs interpretable.

Clinical relevance

Network visualization is used to make sense of relationships in social networks, biological pathways, infrastructure, and knowledge graphs; choosing the right representation and layout helps analysts spot communities, hubs, and pathways that would be invisible in raw relational data.

History

Graph drawing developed strong algorithmic foundations in the 1980s and 1990s, summarized in Di Battista and colleagues' 1999 text. Herman, Melancon, and Marshall surveyed graph visualization for information visualization in 2000, and later controlled studies, such as the node-link versus matrix comparison, refined guidance on representation choices for networks of different sizes.

Debates

Node-link diagrams versus matrix representations for large graphs

Node-link diagrams are intuitive and widely preferred but become cluttered as graphs grow dense, whereas matrix representations remain readable at scale yet are less natural for path-following; the better choice depends on graph size, density, and task.

Key figures

• Ivan Herman
• Giuseppe Di Battista
• Peter Eades
• Roberto Tamassia
• Jean-Daniel Fekete

Seminal works

• dibattista1999
• herman2000
• ghoniem2005

Frequently asked questions

What is a force-directed layout?

A force-directed layout treats a graph like a physical system, with edges acting like springs that pull connected nodes together and nodes repelling each other. Letting the simulation settle produces a layout where tightly connected groups cluster and the overall structure becomes visible, without manual placement.

Why use a matrix instead of a node-link diagram?

For large, dense networks, node-link diagrams become a tangle of crossing edges. A matrix representation places nodes along rows and columns and marks each connection in a cell, avoiding edge clutter and making tasks like spotting highly connected nodes or dense subgroups easier at scale.