Since sprouting up on the scene more than a year ago, neural networks have quickly rooted themselves among credit card issuers.
But as this cutting-edge, pattern-recognition technology becomes more commonplace and widespread, software vendors are venturing deeper into other areas within the rich financial market with new product applications.
Neuraltech Inc., a year-old software company specializing in neural networks, has introduced its latest product for credit card acquiring-side banks and processors to score their merchant customers. The product was to be officially introduced at the American Bankers Association National Operations and Automation Conference, which is taking place in San Francisco this week.
Said to Work Like a Brain
Neuraltech's new neural network system, called Sherlock, has been designed to recognize typically fradulent patterns among merchants and score the account based on a number of variables.
Neural networks can be trained by example to discern patterns, and learn new ones. They are said to work like a human brain. The growing, billion-dollar problem of credit card fraud has heightened interest in these detection methods, creating a need for them outside of just the issuers.
40% Fraud Cut Foreseen
"From a strategic standpoint, it would not seem to make a lot of sense to put all our eggs in one basket," said James Hope, the national sales manager for Neuraltech. "We have a strong track record in merchant scoring, but we're not as strong or differentiated on the issuer side."
Sherlock reviews merchants' profits, volumes, chargebacks, regional location, and past business relationships for risk, looking for patterns within the merchant category and the specific account itself. This neural network system promises to reduce fraud among its users by as much as 40 percent, Mr. Hope claimed.
Currently, he added, Neuraltech has four customers using Sherlock, including a money center bank that he declined to name.
Neuraltech, based in Fairfax, Va., was established with the acquisition last year of CKI, a privately held, 10-year-old software company, and BRNULI, a neural network development firm.
The company currently offers two other products. Cadre, an automated chargeback dispute system for merchant processors, uses expert systems rather than neural networks. The company's Rodin product -- used by Citicorp -- does the same for issuers.
Combination System in Works
Neuraltech is also developing a hybrid system -- a combination of neural technology and expert systems -- for use in the application process on both the merchant and issuing sides. This system should be ready for rollout within four months, Mr. Hope said.
Even with the explosion of neural network technology, Mr. Hope said that his company believes that plenty of room still remains within the financial industry for new neural-based products.
"Banks are more willing to apply this kind of technology to problems," Mr. Hope said.
