Investigation of credit cards fraud detection by using deep learning and classification algorithms

Abstract

Criminal financial behaviour is a problem for both banks and newly created fintech companies. Credit card fraud detection becomes a challenge for any such company. The aim of this paper is to compare ability to detect credit card fraud by four algorithmic methods: Generalized method of moments, Knearest neighbour, Naive Bayes classification and Deep learning. The deep learning algorithm has been tuned to select key parameters so that fraud detection accuracy is the best. Five recognition accuracy parameters and a cost calcualtions showed that the deep learning algorithm is the best fraud detection method compared to other classification algorithms. A financial company reduces losses and increases customer confidence by using fraud prevention technologies.