A few years ago, a prominent data scientist, Chandrasekar Venkatesh, analyzed weather anomalies to prevent road accidents on cable-stayed bridges. He now helps banks and financial institutions address fraud by identifying unusual consumer behavior.
As part of a research team at the University of Cincinnati, Venkatesh developed a tool that alerts drivers to severe weather conditions on a landmark bridge in Toledo, Ohio, that could lead to accidents. This tool has helped save thousands of dollars in public damages from unnecessary bridge closures.
As it turns out, Venkatesh’s lessons apply to the financial services industry, where many banks and companies are struggling to prevent fraud. By detecting anomalies, like cable-stayed bridges, Venkatesh can tackle rising cybercrime, credit card fraud, and more. How can data science help prevent financial threats?
Digging deep into large amounts of data
In his research on traffic road accidents, Venkatesh has to analyze and process data from multiple sources, such as weather sensors, local weather stations, airports, and more, to understand which conditions may lead to adverse driving conditions. This experience has enabled him to successfully prevent debit and credit card fraud.
Detecting anomalies in large amounts of data is critical to identifying criminal activity. Venkatesh said that of the roughly 1 million transactions that local banks make on an average day, 10 could be fraudulent.
“Crime has changed dramatically in the last five or six years because of the introduction of chip cards,” Venkatesh said. “Previously cards only had a magnetic stripe on the back, which was easy to counterfeit. Criminals used to install cameras on ATMs to get the information.”
In 2015-2016, a lot of fraud was eliminated just by using chip cards. But criminals moved online and focused on e-commerce. “Card information can be stolen on bad websites or through data breaches,” Venkatesh said.
For example, this summer, hotel group Marriott International confirmed a data breach that exposed guests’ credit card information. Hackers use social engineering to trick employees into giving them access to computers.
“Mastercard, Visa, Amex, PayPal — they all have algorithms and programs to prevent fraud, but criminals are one step ahead,” Venkatesh said. “Fraudsters use bots to attempt thousands of transactions per minute by guessing card information. That’s why we use data science to stop them.”
Develop risk scores and models
Most fraud prevention and protection models in banking are based on customer transaction patterns. All transactions, spending patterns, average checks and number of purchases per day are part of a consumer’s digital profile.
To develop risk models, data scientists such as Venkatesh analyze the information and share their recommendations with the bank’s software engineers. Risk must be estimated in seconds. The time between when a consumer inserts a card and when the financial institution arrives at a risk score and when a decision must be made to approve the transaction is extremely short.
“If you use your credit card at the grocery store you frequent, the risk score is low,” Venkatesh said. “But, say, your average check at the same store is $40-50, and one day you suddenly buy $500. Then the risk score goes up.”
Data scientists study thousands of variables in transactions. Choosing the right variables is key to detecting fraud. “This is similar to the analysis of weather patterns that I’ve worked on before,” Venkatesh said. “There are thousands of data points, and there are different ways to identify important data points.”
While the way data is collected and analyzed is similar, risk management is different for banks and bridges. For example, some banks enable more transactions during the holidays. They accept some potential fraud-related financial losses in order to provide a smooth customer experience for their regular customers.
“Bridges are not allowed to ‘trade more’ in bad weather conditions to generate more profit as this leads to accidents,” Venkatesh said. “However, if a data scientist makes a mistake in fraud detection, an organization can lose millions of dollars a day. So the stakes are pretty high here as well.”
Image via Shutterstock
This article contains sponsored advertising content. This content is for informational purposes only and does not constitute investment advice.