Determining of credit card frauds: A comparison of artificial neural network and logistic regression methods

Credit card frauds have caused millions of dollars loss each year and exposed the security weaknesses in traditional credit card processing system. Leading banks have been looking for more efficient techniques to determine and prevent credit card frauds. Smart cards are the product of a new technology, which helps to increase the security of the payment transaction. The technology uses a chip system to store transaction data. Even smart cards are more secure compared to magnetic cards; they can not guarantee the determination of fraud card transactions. Previous studies on financial risk assessment have mostly used artificial neural networks, which is a branch of artificial intelligence. The objective of this study is to explore the performance of determining credit card frauds using two popular techniques: Artificial neural networks and logit. These techniques have been used with the data that contains 26 types of TVR (Terminal Verification Results) bites stored during the payments. Both techniques have had significant success; however none of them have stepped forward.