Fraud detection at a financial institution using unsupervised learning and text mining

Description

A recent money laundering scandal at ING resulted in a $900M fine for the bank. David Dogon dives into a best practice use case to demonstrate that for financials, robust and cutting-edge systems to monitor (sometimes millions of) clients are an absolute must. Traditional monitoring systems fall short of adapting to changing behavior and cannot handle complex nonlinear relations in client data, but not having the right systems in place can be detrimental to the business. Join in to learn how machine learning models can provide a solution in cases where traditional systems fall short.

David explains how Van Lanschot Kempen created a system to detect suspicious behavior using unsupervised machine learning techniques on many different sources of internal and external information (including both structured and unstructured data). The result is a more robust and adaptive client monitoring system that also allows fraud analysts to work more efficiently. Along the way, David covers the challenges faced and how Van Lanschot Kempen effectively crossed-checked the machine learning models with external data sources.