Centrality Measures
Centrality measures help us identify the most important or influential nodes in a network. Different centrality measures capture different aspects of importance.
Interactive Lab
Formula Reference
Intuition: Simply count how many connections a node has.
Interpretation: High = well-connected hub. Low = peripheral.
Intuition: Degree as a fraction of all possible connections.
Intuition: Your importance depends on the importance of your neighbors. Connected to important nodes? You're important too.
Intuition: Like eigenvector centrality, but every node gets a free baseline β. The parameter α controls how much neighbor importance matters.
Intuition: Imagine a random surfer clicking links. PageRank measures how often they'd visit each page. α is the probability of following a link (vs. jumping randomly).
Where σst = total shortest paths from s to t, σst(v) = those passing through v
Intuition: How often does this node sit on the shortest path between others? Bridge nodes that connect communities score high.
Intuition: How quickly can this node reach all others? Central nodes can spread information fast.
Why Different Centralities Disagree
| Scenario | High Centrality | Low Centrality | Why |
|---|---|---|---|
| Hub node in star graph | Degree, Closeness | Betweenness (for leaves) | Center connects to everyone but doesn't lie between leaf pairs |
| Bridge node between communities | Betweenness | May have low degree | Only path between groups flows through it |
| Node connected to important nodes only | Eigenvector | Degree (few connections) | Quality over quantity |
| Node with many but isolated connections | Degree | Eigenvector | Connected to unimportant nodes |
Key Insight: Choosing the right centrality depends on the question you're asking about the network. Are you looking for hubs? Bridges? Information spreaders? Each centrality measure captures a different aspect of importance, and they often disagree on which nodes are most central. Understanding these differences helps you choose the right measure for your analysis.