Machine Learning Times
Machine Learning Times

2 years ago
Wise Practitioner – Predictive Analytics Interview Series: Brian Duke at Experian


By: Eric Siegel, Program Co-Chair, Predictive Analytics World for Financial

In anticipation of his upcoming conference presentation at Predictive Analytics World for Financial Las Vegas, June 16-20, 2019, we asked Brian Duke, Data Modeling Director at Experian, a few questions about their deployment of predictive analytics. Catch a glimpse of his presentation, Earn Their Trust: How to Explain Predictive Models to Managers and Predictions to Customers, and see what’s in store at the PAW Financial conference in Las Vegas.

Q: In your work with predictive analytics, what behavior or outcome do your models predict?

A: Our work focuses on assisting our customers in stopping fraud attacks.  Our clients include lenders, tax agencies, retail stores and many other companies interested in verifying consumer identities. We provide them scores, data, and consulting to help them stop fraudsters (ID Thieves, intentional misusers of credit, synthetic identities) more effectively. 

Q: How does predictive analytics deliver value at your organization – what is one specific way in which it actively drives decisions or operations?

A: Our entire business model in Experian Decision Analytics is based on delivering analytic solutions to our customers to enable them to make more effective decisions using data. Despite the recent interest in “machine learning” and related techniques, similar algorithms have been used for ~25 years in financial fraud prevention.

Q: Can you describe a quantitative result, such as the predictive lift of your model or the ROI of an analytics initiative?

A: Fraud detection is a needle in a haystack type of problem. By partnering with our thousands of financial services, retail, and government customers around the world, we have been able to demonstrate the value of our innovations by allowing them to focus their attention on only the riskiest of transactions (<< 1%). Recently, we have been experimenting with some newer techniques such as extreme gradient boosting (XGB) and have seen some clients experience false positive rates 40% lower than with traditional techniques, even when operating on a very small manual review rate.

Q: What surprising discovery or insight have you unearthed in your data?

A: Fraudsters are constantly changing their methods to take advantage of companies.  We need to constantly innovate and learn from data in order to stay ahead of the criminals. Sometimes this means training models more frequently, but we have found the real key to be designing better features and updating those features frequently over time.

Q: Sneak preview: Please tell us a take-away that you will provide during your talk at Predictive Analytics World.

A: Predictive models and artificial intelligence systems greatly impact people’s lives. Headlines about real problems ranging from “fake news” on social media to “algorithmic bias” in systems responsible for personnel decisions and even criminal sentencing make people justifiably suspicious. It is hard to trust that computers are making sound decisions. When models affect people—such as their ability to open a financial account—it is important to explain the algorithm’s reasoning. Explainable AI techniques allow data scientists to use even advanced techniques such as ensemble models while ensuring fairness and transparency.


Don’t miss Brian’s presentation, Earn Their Trust: How to Explain Predictive Models to Managers and Predictions to Customers, at PAW Financial on Wednesday, June 19, 2019 from 11:20 AM to 12:05 PM. Click here to register for attendance.

By: Eric Siegel, Program Co-Chair, Predictive Analytics World for Financial

This content is restricted to site members. If you are an existing user, please log in on the right (desktop) or below (mobile). If not, register today and gain free access to original content and industry news. See the details here.

Comments are closed.