Novel learning strategy based on genetic programming for credit card fraud detection in big data

Abstract

Due to the growing volume of monetary transactions on the internet, credit card fraud poses many challenging issues for banks and financial institutions, thus forcing them to continuously improve their fraud detection systems. However, current fraud detection techniques are far from accurate and fail to minimize the false alarm rates, which can cause inconvenience and dissatisfaction for customers in e-banking system. Furthermore, there are several factors that decrease the performance of Fraud Detection Systems (FDS), such as skewed distribution, concept drift, supports real time detection, large amount of data etc. In this paper, we address the highly imbalanced class issue faced by a FDS when working with big data and propose a novel method of generation of data set's minority class based on K-means clustering method and genetic algorithm to improve classification performance in credit card fraud detection. In our experiments, we apply the proposed approach to synthetic unbalanced credit card fraud data set to demonstrate its effectiveness by means of the most appropriate performance measures for fraud detection purposes.