Overview#
A financial investigator opens a case involving a suspected money laundering operation. The transaction data alone shows nothing unusual: dozens of accounts, hundreds of transfers, no obvious pattern. Then the graph algorithms run. Betweenness centrality immediately flags three accounts sitting at the intersection of every major fund flow. Community detection groups 47 accounts into tight clusters that span seven jurisdictions. Suddenly, what looked like noise resolves into a structured laundering network with clear coordination points.
That is what the Graph Algorithms Library makes possible. It delivers 40+ production-grade algorithms for network analysis, community detection, centrality measurement, and structural pattern recognition. Investigators, data scientists, and intelligence analysts can extract deep insights from interconnected data across financial crime, organised crime, terrorism financing, corporate fraud, and supply chain risk without writing a single line of custom analysis code.
Key Features#
- Comprehensive suite of 40+ algorithms covering centrality, community detection, path analysis, and similarity
- Fast execution of sophisticated algorithms for pattern detection and network analysis
- Automated analysis that accelerates pattern discovery and reduces manual investigation time
- High accuracy in community detection and clustering compared to manual classification
- Incremental algorithms updating results as graphs change without requiring full recomputation
- Seven algorithm categories: centrality measures, community detection, similarity algorithms, path analysis, link prediction, triangle counting and clustering, and connected components
- PageRank and personalized PageRank for influence and importance ranking
- Betweenness, closeness, and eigenvector centrality for network position analysis
- Louvain, label propagation, and modularity optimisation for community detection
- Jaccard, Adamic-Adar, and cosine similarity for entity comparison
- Shortest path, all-pairs shortest path, and widest path algorithms
- Automatic scaling for graphs with hundreds of thousands of nodes
Use Cases#
- Financial Crime Detection: Identify fraud rings, money laundering networks, and coordinated criminal activity through centrality analysis and community detection
- Cybersecurity Threat Mapping: Map threat actor networks, detect attack patterns, and identify critical infrastructure nodes through network analysis
- Intelligence Operations: Analyse terrorism networks, espionage activities, and organised crime hierarchies through sophisticated graph algorithms
- Social Network Analysis: Measure influence, detect communities, and identify key opinion leaders across social platforms
Integration#
- Connects with the Neo4j graph analysis layer and the PostgreSQL primary store through typed APIs
- Supports real-time streaming updates for dynamic network analysis
- Compatible with investigation platforms and case management systems
- Export capabilities for analysis results to visualisation tools and reporting platforms
- Multi-tenant isolation ensuring secure analysis across organisational boundaries
- Horizontal scaling for processing large-scale enterprise graphs with 150,000+ nodes in WebGL visualisation
Open Standards#
- GraphQL (June 2018 Specification): All graph queries, mutations, and real-time subscriptions are exposed through a typed GraphQL API, enabling investigation platforms and tooling to consume graph algorithm results using the standard query language.
- GEXF 1.3 (Graph Exchange XML Format): Graph data is exported in the GEXF 1.3 open format, allowing results to be loaded directly into compatible visualisation and analysis tools such as Gephi.
- RFC 8259, JSON: All node and edge properties, algorithm result payloads, and provenance records are serialised as JSON, the interchange format used across the full API surface.
- RFC 7519, JSON Web Token (JWT): Every algorithm endpoint is protected by RS256-signed JWTs, enforcing authenticated and tenant-scoped access to all graph operations.
- ISO 8601, Date and Time: Timestamps on provenance records, merge operations, and edge temporal properties are serialised in ISO 8601 format, ensuring interoperability with downstream case management and reporting systems.
- Louvain Modularity Optimisation (Blondel et al., 2008): Community detection implements the published Louvain algorithm with configurable resolution parameter, the de facto standard method for large-scale graph partitioning used across academic and industry tooling.
- Jaccard Similarity Coefficient / Adamic-Adar Index: Link prediction and node similarity use these two published, widely standardised graph-theoretic measures, enabling results to be reproduced and compared against other network analysis platforms.
Last Reviewed: 2026-02-05 Last Updated: 2026-04-14