At Capital One, Enhancing Fraud Protection With Machine Learning

Protecting Customers While Reducing False Positives

Capital One is one of the largest banks in the United States, and the largest digital bank. As consumers continue to forgo the brick-and-mortar for the digital-first, Capital One has embraced new technologies, adopting and applying AI and machine learning solutions to nearly every facet of the business and infusing the customer experience with intelligence.


One area where Capital One is applying machine learning is in the realm of fraud detection. Some of the world’s worst cyber criminals focus on the financial services industry, making security all the more vital. According to a 2018 report from the White House council of economic advisors, malicious cyber activity cost the economy between $57 billion and $109 billion in recent years, with the financial sector seeing the most breaches of any industry.

With large masses of data distributed across numerous storage centers, machine learning has been invaluable as Capital One works to better protect its customers’ financial well-being—helping them become more financially empowered, look out for fraud and work to reduce false positives, and better manage their spending.

“We’ve recognized over the past number of years the importance of leveraging machine learning to enhance the user experience, as well as to help us make more informed decisions around engaging with our customers,” says Nitzan Mekel-Bobrov, Ph.D., managing vice president of machine learning at Capital One. “We’re continuously building more sophisticated systems that can leverage a variety of data, structured and unstructured. This allows us to make more precise predictions around whether an activity is fraudulent or not.”

Leveraging a broad suite of machine learning tools and frameworks, such as TensorFlow on Amazon Web Services (AWS), Capital One has the ability to analyze large sums of data, which helps it detect and prevent fraud in real time. When suspicious activity occurs, Capital One will automatically alert customers, walk them through fraud reporting steps, help them lock their card and order a new one, and unlock a temporary card so there is no interruption in their ability to spend their money. With massively more data, longer data history and advanced algorithms, Capital One is focused on leveraging machine learning to revolutionize how it manages fraud.

"We’ve recognized over the past number of years the importance of leveraging machine learning to enhance the user experience, as well as to help us make more informed decisions around engaging with our customers."

Nitzan Mekel-Bobrov, Ph.D.
Managing Vice President of Machine Learning
Capital One

"We’ve recognized over the past number of years the importance of leveraging machine learning to enhance the user experience, as well as to help us make more informed decisions around engaging with our customers."

Nitzan Mekel-Bobrov, Ph.D.
Managing Vice President of Machine Learning
Capital One

For Capital One, using data analytics and machine learning on AWS brings a host of new opportunities to customize their users’ experience and gain more insight on engaging with customers, as well as helping inform key business decisions. This even applies to the bank’s call centers, where speech recognition technology is being used in an automated training and validation system that has roughly doubled accuracy in the past 18 months.

“With machine learning, we’re protecting our customers by preventing fraud. But at the same time, it’s a situation where there are two sides of the same coin,” Mekel-Bobrov says. “On the one hand, this is an essential component of our defensive strategy. But on the other, it’s preventing customers from having a negative experience where they’re being declined when they shouldn’t be. It’s helping us be protective, but not overprotective.”

To Mekel-Bobrov’s point, false positives are well known for irritating and even alienating customers. “What we’re able to do now with machine learning is to continually get better at balancing the two sides of the equation much more dynamically,” Mekel-Bobrov says. “We can optimize on that sweet spot of offering sufficient protection, but not overdo it with too many false positives.”

Crucially, the AWS cloud allows Capital One to deploy a range of both in-house software and machine learning tools, enabling the bank to harness its data in real time and deliver the rapid solutions that are essential in a highly regulated industry. And because the AWS cloud is just as secure—and oftentimes more secure—than an on-premises data center, it’s able to apply these innovations while upholding its responsibility to protect customers and their data.

“With AWS and our move to the cloud, we can build a truly modern machine learning ecosystem, with all of our data connected and fully available,” Mekel-Bobrov says. “This allows us to deploy models that are automatically set to incoming data, automatically scaling our infrastructure and even plugging in our own solutions for added flexibility.”—so that we can focus on a potentially life-saving use of machine learning—the better.”

From partners.wsj.com

Kia uses ML to reduce fatalities from auto accidents

Learn more »

Zendesk helps companies provide faster customer service using ML

Learn more »

GE Healthcare drives better healthcare outcomes with ML

Learn more »

T-Mobile humanizes customer service using ML

Learn more »