Fighting Fraud with Graph Databases

Explore graph databases for fraud detection in this 42-minute conference talk from Devoxx. Learn how graph databases excel at representing complex relationships between entities, outperforming traditional SQL databases in both response time and query writing. Discover the basic concepts of graph databases and gain insights from real-world experience in fraud detection within the Belgian social security sector. Understand the advantages of graph databases in various domains, including recommendation systems, social networks, and master data management. Follow along as the speaker introduces graph database models, demonstrates Neo4j requests, and compares implementation approaches with SQL/RDBMS. Examine practical examples such as Limosa detachment declarations and sub-contractor relationships, while also addressing data quality challenges like duplicate entries and address matching.

Intro
Outline
Motivations
Starting points
RDBMS and relationships: limitations
Graph database model
GraphDB: objectives
GraphDB : less code
Neo4j request
How it works
Limosa: detachment declaration
DDT: sub-contractors
Full schema
Full query
Implementation
In SQL/RDBMS ?
Data quality problem
Addresses
Duplicates combinations
Schema with B duplicated
Conclusions
References : www.smalsresearch.be