• Customer Risk Rating
  • Financial Crime
  • Anti-Money Laundering
  • Artificial Neural Networks
  • AML
  • ANN
June 28, 2017

Machine learning for customer risk ratings

Financial crime and compliance analytics
 

Financial institutions are mandated to conduct anti-money laundering (AML) risk assessments to determine the overall risk rating of their customers. The assessment and subsequent rating are derived using a risk rating methodology based on industry standards and rules defined by regulatory bodies. Customer risk rating is a typical multi-class problem because financial institutions have two or more categories of ratings. Most firms highlight the importance of judgment/qualitative factors in determining the risk rating of customers. However, in practice, a mathematical model using quantitative variables/attributes plays an important role in determining customer risk ratings because the process is convenient and consumes less time. This article discusses the application of machine learning for customer risk ratings.
 

Objective
 

This document proposes a framework/methodology based on machine learning approach to establish the risk rating score (normally a low, medium or high score) of customers using various drivers/attributes as input variables. The methodology is used for AML risk assessment and rating during the onboarding phase, and the procedure will be performed on an ongoing basis throughout the period of the customer’s association with the financial institution.