A Classi cation System for Credit CardTransactionsEliseo

This paper describes a system that responds to a signiicant problem involving credit card theft. After a card has been stolen, some of the transactions processed subsequently have to be granted if it seems that they were carried out by the card owner himself (e.g. by a standing order) while other transactions should be denied authorization if they were carried out by the misuser of the card. The system developed makes use of a neural network integrated with a Case-Based Reasoning (CBR) component. While the CBR system can use speciic cases to determine when a transaction should be granted or denied authorization, the neural network can recognise general patterns of behaviour for the use and the misuse of credit cards and can use this knowledge in the classiication task. This paper presents the architecture of the classiication system proposed, highlighting the main aspects of each of its components. The system's performance is evaluated and conclusions are drawn.