Classification and anomaly detection

Anomaly detection

Anomaly detection surfaces records that deviate from learned patterns, which makes it a triage tool for reconciliation, transaction review, and data quality.

Operational use
Flagging unusual transactions, duplicate candidates, drifting data feeds, and reconciliation differences worth a person's attention first.
When not to use it
When the definition of wrong is precise and known, a rule is better than a model. Anomaly scores should never auto-reject records.
How it is validated
Precision on a reviewed sample: of the records flagged, how many were worth flagging. Thresholds tuned to reviewer capacity, monitored over time.
Limitations
An anomaly is a statistical statement, not an accusation. Detection surfaces unusual records; it does not independently prove fraud, error, or misconduct.

Isolation Forest, the reference method here, works on a simple insight: anomalous points are easier to isolate than normal ones. It is fast, robust, and needs no labeled fraud examples, which suits operational data where labels rarely exist.

In Eigenvalue systems, anomaly detection is a prioritization layer: it decides what a reviewer sees first, never what gets rejected. The score arrives with the evidence beside it, and the reviewer’s decision feeds back into the record. The method’s honest job is making a large review queue small, not replacing the review.

Where it appears in our work

Primary sources

Last reviewed 2026-08-02. Sources link to primary or official pages; verify current versions before relying on them.