Credit Card Fraud Detection Based on Multilayer Perceptron and Extreme Learning Machine Architectures

Abstract

Due to the increasing digitalization of banking services and the predominance of mobile banking applications, the rate of credit card payments is increasing every year, among billions of transactions identified as fraudulent. Data mining algorithms have played a fundamental role in detecting fraudulent transactions, through combating fraudster's attacks working around classical fraud prevention systems. In this paper, we try to detect fraudulent transactions using two artificial neural network classifiers, Multilayer Perceptron (MLP) and Extreme Learning Machine (ELM), applied on the credit card fraud dataset. The performance of these classifiers is evaluated based on accuracy, recall, precision, and classification time. The results show that the accuracy of MLP and ELM classifiers achieves respectively 97.84% and 95.46%. Otherwise, ELM is very fast for predicting new fraudulent transactions.