Models / Anti-Money Laundering

AML Alert Prioritiser

Ranks transaction-monitoring alerts by likelihood of an STR.

Strictly ConfidentialTier 1 modelFinancial crimeBanking secrecy+1High riskAnomaly Detection · 2.1.0 · aml-alert-prioritiser
34 27K/30d 47

Precision@top 10% of alerts (STR filed)

0.37

Recall of STR-filed alerts in top 40%

0.96

Alert reduction (light-review band)

44%

Parity vs SAS champion (AUC delta)

+0.021

About this model

Scores every alert raised by the transaction monitoring system on how likely it is to end in a suspicious transaction report, so investigators work the riskiest alerts first and low-risk alerts get a lighter review. Each score carries reason codes (structuring pattern, high-risk jurisdiction counterparty, unusual cash activity for the profile).

Intended use

Queue ordering and review-depth routing for Financial Crime Compliance investigators. No alert is closed by the model: low-scored alerts still get a documented human review, and sampling checks the low band every month.

Training data lineage

Rebuilt from the SAS Enterprise Miner alert-scoring model used by Financial Crime since 2021. Trained on three years of dispositioned alerts in aml-alerts-history, with customer activity features from casa-transactions, in BigQuery and Vertex AI. Tier 1 model validated by Model Risk Management; below-the-line testing and champion-challenger against the SAS model ran for two months before switch-over.

Limitations & bias notes

Learns from past investigator decisions, so any historical under-reporting pattern can carry through; the low band is sampled monthly to catch this. New typologies need new monitoring scenarios first; the model does not raise alerts itself.

#aml#transaction-monitoring#alert-triage#financial-crime#reason-codes#tier-1#rebuilt-from-sas

Ownership and sensitivity

Owned by

Group Financial Crime ComplianceAnti-Money Laundering

Accountable owner: Ravi Chandran

Who approves access

  1. Owner, Anti-Money Laundering, Ravi Chandran
  2. Data Governance Office, Ahmad Faizal
  3. Model Risk Management, Dr. Kavitha Subramaniam
Strictly Confidential

Customer-level data under banking secrecy and PDPA, and models that decide on customers.

Named-user entitlement with the data owner's consent and a stated purpose; reviewed on expiry.

Tier 1 model
Material model: independent validation before use and annual revalidation.
Financial crime
Used in AML, fraud or sanctions controls; tipping-off restrictions apply.
Banking secrecy
Customer account information protected under the Financial Services Act 2013.
Customer PII
Contains or processes personal data about customers (PDPA 2010).

Try it

Live sandbox

alert=AML-2026-118204&scenario=cash-structuring

fixture

Evaluation only: restricted models run against synthetic or masked sample data until your business unit's access request is approved.

Feedback

Details

Updated
2026-07-17
Latest version
2.1.0
Licence
BU Restricted
Access
Restricted
Framework
XGBoost
Language
N/A