COMPLIANCE & FRAUD
Anti-money laundering pattern detection: read the network, not just the transaction
Anti-money laundering pattern detection with TIMi looks for unusual combinations or sequences of transactions that may point to an attempt to hide the origin of funds, and shows the factors behind each alert.
Network view
Flows between accounts are analyzed as a network, not one transaction at a time.
Unusual structures
Split amounts and rapid back-and-forth between accounts can be surfaced.
Explained per alert
Each alert is documented factor by factor.
HOW IT WORKS
How anti-money laundering pattern detection works.
Three steps, in one engine, on a standard server with no cluster.
1. Prepare the data
Anatella gathers the transaction history, the relations between accounts or entities and the typologies already documented, then structures them into variables.
2. Map the network and model
Anatella’s graph mining module (social network analysis) represents flows between accounts as a network to spot unusual structures. Modeler then estimates how likely a sequence is to match a risky pattern.
3. Score and explain
Each transaction, account or group of accounts gets a suspicion level with its factors, such as the structure of the transaction network or a gap with the usual behavior of the account.
Anatella
Graph mining and ModelerSCOPE
What TIMi does, and what stays with you.
Anti-money laundering is governed by precise regulatory obligations, which vary by jurisdiction and by sector (banking, insurance, gaming, real estate).
A technical capability, not a compliance program
TIMi provides the technical capability to detect unusual patterns from the data you hold. Compliance with specific obligations (reports to authorities, legal thresholds, mandatory typologies) remains the responsibility of your organization and its compliance teams.
A complement to your existing tools
The approach adds pattern detection to the rules and thresholds already in place. It does not replace a complete regulatory compliance system.
With or without internal cases
A history of confirmed cases improves precision. Without enough internal cases, the approach can rely on documented typologies and general anomaly detection.
EXPLAINABILITY
An alert says suspicious. The factors say why.
An alert alone says an account or a transaction looks suspicious, not why. A white-box model shows the factors behind it, so compliance teams can document their analysis and justify a decision to investigate or to report.
For compliance teams
Each alert can be documented factor by factor, which helps the traceability of decisions.
For model governance
An explainable model can be simpler to audit and to have validated by an internal control function.
For model owners
Typologies evolve. Modeler’s built-in Drift Report shows when a model loses accuracy, so you update it when needed instead of on a fixed calendar.
DATA
What data an AML pattern model uses.
Anatella connects to more than 80 sources, so this data can be gathered without prior custom development.
Transactions
History of transactions between accounts or entities.
Relations
Identification data and links between accounts.
Documented typologies
Patterns already described or observed.
Thresholds and rules
Existing rules, integrated alongside the model.
PROOF
Closest references.
We do not yet have a published customer story on money laundering detection. These two deployments show building blocks: processing very large transaction volumes, and detecting patterns in national databases.
150 billion transactions a month on one server
Anatella processes the monthly transactions of a national telecom network, including fraud control, on a standard server in 5 to 7 hours.
Criminal patterns detected in national databases
Text mining applied to national databases, with data refreshed in under 6 hours each day.
FAQ
Common questions.
Does this replace a dedicated AML compliance tool?
No. It adds a pattern detection capability from the data you hold, next to the rules and thresholds already in place. It does not replace a complete regulatory compliance system.
How does network analysis help?
Some laundering schemes show up as a structure of transactions between several accounts rather than as one isolated transaction. Representing it as a network reveals unusual configurations that a transaction-by-transaction analysis would miss.
Do we need confirmed laundering cases to train the model?
A history of confirmed cases improves precision, but the approach can also rely on documented typologies and general anomaly detection when internal cases are not sufficient.
Does it work in real time?
Processing can be set up in near real time or in batches, depending on operational needs and the regulatory constraints that apply.
SEE FOR YOURSELF
Test it on your own data.
Run TIMi on your own data with a free trial, or talk to us about your case.
