Equifax
Powering Innovation with Machine Learning and Explainable Artificial Intelligence

Credit lending is a massive business domain in the U.S., directly and indirectly influencing almost all areas of the economy. With a major share of the population holding loans worth trillions of dollars, any technology capable of making even a small advance in a financial organization’s bottom-line or market-share would be an opportunity for substantial growth. For that reason, both established banks and fintechs are constantly looking for ways to innovate. Both Artificial Intelligence (AI) and Machine Learning (ML) present excellent opportunities to expand portfolios. At its core, lending is a big data challenge, making it a function perfectly suited for ML. The more data an organization can leverage well, the better it can increase business and gain customer traction while eliminating administrative overhead, bad decisions and delays. AI and ML come with a promise to analyze multiple data sources and models to deliver optimal decisions aimed at improving outcomes for consumers and for businesses.

In this regard, Equifax is a powerhouse that is proactively applying groundbreaking modeling techniques fueled by explainable AI (xAI) and ML in the credit decisioning landscape. In 2015, Equifax developed a patented solution, NeuroDecision Technology^®, powered by xAI, to help lenders approve more consumers for credit without adding risk to a portfolio. The technology is among the first ML credit scoring methodologies to provide explainable reason codes for consumers. Today, Equifax is working toward a launch of their fourth generation of NeuroDecision technology.

To understand the latest trends in xAI and ML modeling, CIO Applications Magazine interviewed leaders from Equifax: Christopher Yasko, Vice President of the Innovation Lab; Matthew Turner, Fellow and Principal Mathematical Statistician; and Rajkumar Bondugula, Data Science Fellow, who together play a pivotal role in shaping the company’s innovation roadmap.

Could you give us a brief overview of Equifax?

Equifax has a 125-year history, with a current laser focus on leveraging data, analytics and technology to empower businesses to make better, more confident decisions while helping consumers get access to the credit they deserve.

With a mission to innovate for Equifax's future, we have built an Enterprise Innovation Lab inside the Data & Analytics team. At the lab, we explore future business needs and pioneer disruptive solutions that will help drive new insights. The lab employs scientists with PhD degrees in a wide range of academic subjects, including engineering, mathematics, statistics, and more.

We try to solve real-life problems by applying ML, xAI, and data science with our expertise.
What does Equifax aim to achieve through machine learning?

A core focus is the credit Industry, which is a highly regulated segment. With the establishment of the Fair Credit Reporting Act (FCRA), it became essential for organizations to provide consumers with key factors that negatively impact credit decisions. When we started the Innovation Lab, we focused on new solutions to help improve business and consumer outcomes that also explain the key factors that might negatively impact a loan applicant's credit score. We are the pioneers who built a solution that applied ML in FCRA decisions and have patented numerous xAI solutions. By taking these solutions to regulatory agencies in the U.S. and Europe, we gained mindshare around what was possible. Currently, we are working with unstructured data, streaming data, and time series data to further raise the bar.

Our capabilities in ML can be better understood through an example. To run a hierarchical clustering on 200 million people, we need to perform about 40 quintillion computations. Even with all our computational resources, achieving 40 quintillion computation is intractable. Practically, it takes four years to run on the computations on 3000 CPU's, which is not a feasible way. Equifax has developed novel algorithms by rewriting the codes, which enable us to bring down the computations from 40 quintillion to about 70 quadrillion. 70 quadrillion is still a huge number, but it is certainly tractable as we are able to process that 200 million data points in a couple of hours. Our specialty is not just AI and making it regulatory compliant, but running AI at a massive scale. We write and rewrite algorithms and training models across millions of data points to make them compliant with regulatory guidance. It is a highly challenging - and rewarding - task.

What can we expect from the fourth generation of NeuroDecision?

The fourth-generation xAI algorithm comes with advanced computing capabilities and big data processing capabilities, which are facilitated by Cloud technology. They are also designed to consume more types of data. Recently, Equifax has acquired significant alternative and proprietary data assets. Leveraging powerful data, technology and analytics, we seek to open up more growth opportunities for our clients and more ways to access credit for consumers.

What are some of the factors that differentiate Equifax from other players?

Equifax is on a transformative journey. Over the last three years, we have invested $1.5 billion in technology and migrated thousands of customers to the cloud to help them succeed and grow.

We are at the forefront of ML and explainable AI technology. In the last four years, we have obtained more than two dozen patents in these areas. We can process huge volumes of data in the cloud. More importantly, we work across a wide range of data types, many of them differentiated and proprietary to deliver robust and actionable insights.

Our team at the D&A Data Science Lab engages extensively with key customers, industry peer groups, and universities. We have a faculty research sponsoring program that provides opportunities to students to work with us. We work with several colleges in the state of Georgia, including the Georgia Institute of technology, Kennesaw State University, and more. Such initiatives provide opportunities to Ph.D. students to do independent research and make significant contributions to science. We also have teamed up with Cornell University in New York, Stanford School of Economics, and Harvard business school to facilitate research opportunities. Our technical staff has a Fellows program through which we identify the brightest and talented employees of our company who are worthy of the Fellow designation.