Delivering Anti-Money Laundering Predictions with AI

This post was originally part of the DataRobot Community. Visit now to browse discussions and ask questions about DataRobot, AI Platform, data science, and more.

The Problem

Money laundering is essentially the process of trying to conceal the source of ill-gotten funds—illegal activities like tax evasion, drug dealing, human trafficking, and terrorist financing. Launderers pass funds through the financial system in an attempt to make them appear legitimate. Anti-money laundering teams, meanwhile, develop and deploy monitoring programs throughout their institutions in search of behavior consistent with money laundering. Programs that do not meet the required regulatory standards or have significant lapses can incur large fines—sometimes to the tune of hundreds of millions, or even climbing into the billions of dollars. The currently popular rule-based systems are rapidly becoming out of date and are unable to handle the complex and sophisticated tactics of money launderers. In fact, the false positive rates generated by these rule-based system can reach very high levels of more than 90%.

The Challenge and the Solution

Maintaining anti-laundering programs is expensive, partially due to the cost of assigning a large team of investigators. These teams can involve hundreds of reviewers at the larger banks, and for the largest banks, more than a thousand. The personnel cost is dramatic, yet the programs generate no revenue for the business. This environment is a prime motivation for increasing the efficiency of the current transaction monitoring and detection processes. Enter DataRobot’s enterprise AI platform.

DataRobot’s automated machine learning guides the process of selecting a model that most accurately predicts which cases have the highest likelihood of suspicious activity. With the best model identified, the team can deploy it into production to score future alerts. The scores output by the model are then used to rank new cases. Any case exceeding a score threshold is sent for review; cases below the threshold can be discarded or subjected to a lighter review. Applying automated machine learning helps reduce the false positive rate of cases selected for manual review, which translates into greater efficiency for the compliance team.

Example

Shruthi works as an analyst on the internal investigations team at a large bank. Her boss, Katrina, heads up the bank’s financial crimes unit within the bank’s compliance organization. Katrina has been tasked with improving the efficiency of the bank’s anti-money laundering transaction monitoring operations. Today’s rule-based system triggers an alert whenever a transaction satisfies some simple conditions indicating that it is potentially suspicious. The problem is that when these alerts are manually reviewed by Shruthi and her colleagues, it turns out that 9 out of 10 are false alerts. Katrina is implementing DataRobot’s automated machine learning solution for use by her analysts. They can then prioritize which alerts legitimately require manual review, allowing Katrina to dedicate more resources to those cases most likely to be suspicious.

• Training data
• Prediction data
• Data Dictionary

About the author

Linda Haviland

Community Manager

Meet Linda Haviland