A Comparative Study and Performance Analysis of ATM Card Fraud Detection Techniques

Abstract

ATM card fraud is increasing gradually with the expansion of modern technology and global communication. In the whole world, it is resulting in the loss of billions of dollars each year. Fraud detection systems have become essential for all ATM card issuing banks to minimize their losses. The main goals are, firstly, to review alternative techniques that have been used in fraud detection and secondly compare and analyze these techniques that are already used in ATM card fraud detection. Recently different card security systems used different fraud detection techniques; these techniques are based on neural network, genetic algorithm, hidden Markov model, Bayesian network, decision tree, clustering method, support vector machine, etc. According to our survey, the most important parameters used for comparing these fraud detection systems are accuracy, speed and cost of fraud detection. This study is very useful for any ATM card provider to choose an appropriate solution for fraud detection problem and also enable us to build a hybrid approach for developing some effective algorithms which can perform properly on fraud detection mechanism.