Networks show up everywhere: social media connections, transportation routes, payment rails, supply chains, and even interactions between services in a microservices architecture. When we model these systems as graphs, a common question appears quickly: which nodes matter most for how information, traffic, or influence moves through the network? Betweenness centrality is one of the most practical measures for answering that question. It highlights nodes that frequently lie on the shortest paths between other nodes, making them potential “brokers,” “bridges,” or “choke points.” For learners exploring graph analytics in a data science course in Kolkata, betweenness centrality is a core concept because it connects graph theory to real operational decisions.
What Betweenness Centrality Measures
Betweenness centrality quantifies how often a node sits on the shortest paths between pairs of other nodes. If many shortest routes pass through a node, that node can strongly influence (or control) the flow of information.
The intuition in plain terms
Imagine a city map: if a particular bridge connects two large regions, most shortest commutes may cross it. That bridge has a high betweenness. In a workplace communication network, a manager who is the only link between two teams often has high betweenness even if they do not have the most direct connections (degree).
The standard definition (conceptually)
For a node v, betweenness centrality is the sum over all node pairs (s, t) of the fraction of shortest paths from s to t that pass through v.
- If there are multiple equally short routes, each route contributes proportionally.
- Many tools also provide a normalised version so values can be compared across graphs of different sizes.
How It’s Computed and Why Complexity Matters
Computing betweenness centrality naively can be expensive because it involves shortest paths between many pairs of nodes.
Common approach: Brandes’ algorithm
Most modern graph libraries rely on Brandes’ algorithm, which significantly reduces computation time by reusing shortest-path calculations:
- For unweighted graphs, it runs efficiently using BFS-based shortest paths.
- For weighted graphs, it uses Dijkstra-based shortest paths from each source node.
This matters in real projects. If you analyse a network with hundreds of thousands of nodes (for example, clickstream graphs or telecom graphs), exact betweenness may be slow. In those cases, practitioners often use:
- Approximation by sampling a subset of source nodes
- Graph sparsification or filtering
- Parallel/distributed graph engines
In a practical data science course in Kolkata, it’s useful to learn not only the meaning of betweenness but also when approximations are acceptable, because production constraints often matter as much as theory.
Practical Use Cases That Benefit from Betweenness Centrality
Betweenness centrality is valuable when your goal is to identify “bridges” that connect communities or routes.
1) Social and organisational networks
In organisational graphs (email, collaboration, issue tracking), nodes with high betweenness often act as coordinators between groups. This can help with:
- Identifying single points of failure in knowledge flow
- Detecting informal brokers who connect teams
- Planning cross-team communication to reduce bottlenecks
However, high betweenness is not always “good.” It can indicate overload risk: if too much coordination depends on one person or one service, the system becomes fragile.
2) Fraud, risk, and compliance networks
Financial crime analytics often model entities (accounts, devices, merchants) as nodes and transactions as edges. Nodes with high betweenness can indicate:
- Intermediaries linking suspicious clusters
- Mule-account patterns bridging otherwise separate groups
- Critical connectors in laundering paths
This is not a standalone proof of fraud, but it is a strong feature when combined with others like community detection, temporal patterns, and anomaly scores.
3) Transportation, logistics, and network resilience
In road networks, airline route maps, or supply-chain graphs, high-betweenness nodes may be:
- Critical hubs or junctions
- Ports, warehouses, or routes whose disruption causes widespread delays
- Candidate points for monitoring, redundancy planning, or preventive maintenance
A useful exercise often included in a data science course in Kolkata is comparing betweenness with degree: a busy-looking hub may have many direct connections, but a different node may actually be more critical because it lies on key shortest routes between regions.
Interpreting Results Correctly: Pitfalls and Best Practices
Betweenness centrality is powerful, but interpretation must match the context and assumptions.
Shortest-path assumption may not hold
Betweenness assumes flows choose shortest paths. That works well for routing systems or certain diffusion processes, but not always for human behaviour, where information can spread through non-shortest routes.
Sensitivity to graph construction
Your definition of nodes and edges can change the result dramatically. Examples:
- Directed vs undirected edges (followers vs mutual connections)
- Weighted edges (cost, distance, latency, transaction volume)
- Temporal graphs (relationships change over time)
A good practice is to run “stress tests”:
- Compare weighted vs unweighted versions
- Evaluate stability across time windows
- Check whether high-betweenness nodes remain important after removing noisy edges
Use with complementary metrics
Combine betweenness with:
- Degree centrality (local connectivity)
- Closeness centrality (average distance to others)
- Eigenvector/PageRank-style measures (influence via influential neighbours)
- Community detection (to see which communities a node bridges)
Conclusion
Betweenness centrality helps you find nodes that shape how information, traffic, or influence moves across a network by highlighting those that frequently lie on shortest paths. It is especially useful for identifying bridges between communities, pinpointing bottlenecks, and improving resilience in social, financial, and infrastructure graphs. The key is to interpret it carefully—understanding how your graph is built, whether shortest-path assumptions make sense, and when approximations are needed. Mastering these practical details turns centrality from a formula into an actionable insight—exactly the kind of applied thinking learners aim for in a data science course in Kolkata.