PageRank

Find the most influential nodes in your network

PageRank measures node importance based on the structure of incoming connections - nodes are important if they're connected to by other important nodes.

What It Computes

For each node, PageRank calculates an influence score (0 to 1) representing the probability that a random walker ends up at that node after many steps.

Key insight: Quality of connections matters more than quantity. One link from a highly-ranked node is worth more than many links from low-ranked nodes.

When to Use It

✅ Fraud Detection

Identify money mule accounts and fraud ring leaders

✅ Cybersecurity

Find critical nodes in attack infrastructure

✅ Social Analytics

Discover influencers and opinion leaders

✅ Knowledge Graphs

Rank entities by importance

Parameters

Parameter	Type	Default	Description
`iterations`	int	20	Number of iterations to run
`damping`	float	0.85	Probability of following an edge (vs random jump)
`tolerance`	float	0.0001	Convergence threshold

Tuning advice:

More iterations: Better convergence, slower runtime (20-100 typical)
Higher damping: More weight on graph structure (0.85 is standard)
Lower tolerance: More precision, more iterations needed

Performance

Time Complexity: O(E) per iteration
Space Complexity: O(V)
Typical Runtime: 1-2 seconds for 1M edges (20 iterations)

Scales to: 100M+ edges with good performance

Example

Output:

Real-World Use Cases

Financial Fraud Detection

Problem: Identify money mule accounts in transaction networks
Solution: High PageRank = receives money from many sources

Cybersecurity

Problem: Identify C2 servers in botnet traffic
Solution: High PageRank = many infected hosts communicate with it

Problem: Find influencers for marketing campaigns
Solution: PageRank identifies users whose content reaches the most people

Temporal PageRank

For temporal graphs, run PageRank on different time windows to track influence evolution:

Use case: Detect rising/falling influencers, emerging fraud networks

Comparison with Other Centrality Metrics

Metric	What It Measures	When to Prefer PageRank
Degree	Number of connections	PageRank better for quality over quantity
Betweenness	Bridge position	PageRank for general influence
Closeness	Average distance to others	PageRank for directed graphs

Rule of thumb: Use PageRank for influence, betweenness for critical connectors

Performance Tips

Use fewer iterations for large graphs (10-15 sufficient)
Parallelize: Raphtory automatically uses multiple threads
Cache results: Store scores if querying repeatedly
Sample: For exploration, run on graph sample first