In fraud detection, a suspicious account may matter because of its connections, not just its own transactions. Centrality is a value assigned to a node by a measure of its position in a graph. Accounts, devices, and addresses can be nodes; transactions or shared resources can form edges. Centrality helps investigators identify important accounts in networks that transaction-by-transaction analysis may miss. section 7.4.3
Choose a measure according to the investigative question:
- PageRank can highlight influential nodes, such as possible central figures in a fraud ring or key money-laundering accounts. section 7.4.3
- Betweenness centrality can highlight nodes that act as crucial bridges in illicit flows. section 7.4.3
Centrality scores are only as useful as the graph behind them. Validate imported data for orphaned nodes and inconsistent relationships, then add computed scores as node properties or export them to a relational database for reporting and integration. In-memory libraries such as NetworkX or igraph suit exploratory and smaller-scale work; a graph database may better fit a dynamic production system. Neither choice replaces data-quality checks. section 9.2.3 section 9.2.3
