Banks use data mining to beat card fraudsters

A specialist data mining and analytics company SPSS is working with cash point network, LINK in a bid to identify fraudulent...

A specialist data mining and analytics company SPSS is working with cash point network, LINK in a bid to identify fraudulent withdrawals.

SPSS has implemented its Clementine data-mining workbench to monitor withdrawals from LINK machines and identify behavioural or usage patterns of cards in a bid to establish a database of fraudulent activity.

Cash machine fraud currently costs banks £21m a year. In a bid to reduce losses, LINK started working with SPSS in March.

"In March 2001 alone, LINK identified 3,300 suspect card details using the SPSS solution, and from this information 49 cases of fraud were identified with an average saving of £450 per card," Alex Leckenby, security administrator at LINK, told CW360.com.

LINK currently identifies potential fraudulent activity and then contacts customers to check if their card is missing or has been stolen. According to SPSS senior sales consultant Kate Devaney, LINK will now be able to develop a definitive database of fraudulent activity enabling it to refuse requests or take the card away.

"Clementine sits on the [LINK] system and takes a download of up to six million transactions a day, analyses them and processes the results via Oracle Reports," explained Devaney.

Clementine is looking out for specific patterns that could be fraudulent such as withdrawals of large sums just before and just after midnight, use of the card in different locations in a short period of time, and high volume withdrawals from an account where only small sums of money are usually taken.

The data mining software has been installed at LINK's site in Harrogate, Yorkshire.

Read more on Business applications

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.

-ADS BY GOOGLE

SearchCIO

SearchSecurity

SearchNetworking

SearchDataCenter

SearchDataManagement

Close