Many datasets are about relationships: who emails whom, which products are bought together, which accounts transfer money. Network analysis studies these connections.
Key Concepts
- Nodes: the entities.
- Edges: the relationships, which may be directed and weighted.
Centrality Measures
- Degree: number of connections.
- Betweenness: how often a node lies on paths between others — a broker.
- Closeness: how quickly a node reaches others.
- PageRank-style measures: importance based on connections to important nodes.
Community Detection
Algorithms find groups of densely connected nodes — customer communities, fraud rings, topic clusters.
Uses
- Fraud detection: suspicious clusters of linked accounts.
- Organisational analysis: collaboration patterns.
- Recommendations from co-purchase networks.
- Supply chain dependencies.
- Knowledge graphs for search and AI.
Tools
Python libraries such as NetworkX, graph databases and visualisation tools.
Visualisation Caution
Large network diagrams quickly become unreadable "hairballs". Filter, aggregate and highlight.