As fraud-fighting becomes increasingly complicated, machine learning has emerged as a new holy grail technology for fraud detection. In “Machine Learning: Fraud Is Now a Competitive Issue,” 65 percent of the senior fraud and data analytics executives at North American financial institutions Aite Group interviewed say the priority for investment in machine-learning analytics for fraud mitigation is very high and a key area of investment; another 35 percent call it a moderate priority. Yet, only 40 percent of interviewees have an ML-enabling platform deployed or in production, with another 10 percent with proofs of concept underway.
The question for credit unions is, how and when will machine learning make a difference in your fraud-fighting efforts?
Insight Vault caught up with Nick Calcanes, CO-OP’s chief information officer, and Paulo Marques, chief technology officer and co-founder of pioneering AI developer Feedzai to get some answers. Full disclosure: CO-OP is teaming up with Feedzai on a new machine-learning security tool for the credit union industry that is expected to deploy in 2018.
Here are five pointed questions about machine learning, fraud and how your credit union might fit in:
- CO-OP is already doing a lot to combat fraud. What does machine learning bring to the party?
“Machine learning allows us to access and process a great deal more data than traditional methods do,” Calcanes says. “Data scientists have to build models and teach the machine, but once they do, the program learns at a pace that is far beyond what we humans can match. Fraudsters are constantly evolving. Machine learning gives us the speed and flexibility to keep up.”
“It’s almost like giving yourself superpowers,” explains Marques. “Instead of writing rules by hand, the machine itself learns what’s fraud and what isn’t and helps you to fight it at scale and in real time.”
- Why is machine learning necessary today?
Three reasons. First, the sheer volume of transactions has increased substantially. According to the Nilson Report, credit card transactions rose 48 percent, debit card transactions 46 percent and electronic transactions 45 percent between 2010 and 2015, for a collective increase of 34.2 billion transactions annually.
In addition to volume, the variety of transactions is also exploding. Mobile wallets, the IoT, P2P networks, digital banking – all of these contribute to greater complexity. As a result, existing fraud detection has trouble “seeing” the transactions coming in. Data is also disjointed, blurring the level of detail and making it difficult to perceive trends.
Finally, even if you aren’t using machine learning yet, fraudsters are. “Fraudsters are increasingly sophisticated,” says Marques, “and fraud is adversarial. Wherever you create a rule, fraudsters are working on a way to get around it.”
- How will CO-OP deploy machine learning?
Being able to use machine learning is just one of the benefits of CO-OP’s ongoing effort to unify its ecosystem. CO-OP is creating a platform that pulls together its transaction data. This is a necessary step for machine learning to work. As the data is aggregated and the machine learning system reaches full steam, the system will learn as it goes. Initially, machine learning will work side by side with CO-OP’s neural network technology. Over time, CO-OP may switch to machine learning entirely or keep both systems in place as the ultimate safeguard.
“We expect to have our machine learning tool in place in 2018,” says Calcanes. “We then will continue to deploy and tune our models.”
- What makes the CO-OP connection valuable here?
CO-OP is on track to process more than 4 billion transactions this year. This volume of data not only makes machine learning a necessary security enhancement; it also “feeds” the machine learning system by providing the volume of data needed to detect trends and learn effectively.
CO-OP plans to use its scale to maximize the value of new machine learning technology. “We won’t share real data across credit unions, but we will aggregate it for modeling,” says Calcanes. “Whether you’re a $3 billion credit union or a $300 million credit union, our teams will work with you to make sure we’ve aligned the full leverage of this technology to your specific needs.”
Data and scale are huge challenges to implementing this technology, says Marques: “The initial phase [of aggregating data and achieving scale] is the hardest. Data can live in many different systems and silos. Often the most painful part is learning where all the data is and connecting to it.”
Having CO-OP undertake this challenge on behalf of its clients makes access a reality, especially for smaller and mid-size credit unions.
- What will the future look like for machine learning and credit unions?
“Fraud is a hot topic and, because the cost of fraud is so high, it’s a great rationale for investing in this technology,” says Calcanes. “We’re using this initiative to stand the platform up and leverage the capability across all we do.”
These steps toward data aggregation will also help CO-OP and its client credit unions leverage the intelligence that their data holds, potentially opening the door to predictive analytics and other forms of AI.
“Machine learning is probably the most important technology to emerge in the past five years,” says Marques. That credit unions will have the opportunity to put it to work full force – first against fraud, then to improve marketing and member experience – is big news for the industry and its members.”
Want to know more about machine learning and its impact on you and your credit union? Download CO-OP’s free eBook “A New Frontier: Machine Learning, Artificial Intelligence and Big Data.”