Credit card fraud detection using machine learning algorithms

Abstract

In this work, we describe a tutorial to solve the fraud problem under a supervised learning context in machine learning, and this tutorial consists of a set of methodologies that allow the construction of a model for recognizing fraudulent transactions in payments via card credit. For this, firstly, we explain the concept of fraud in means of payment, its consequences, and the importance of recognizing this type of transaction for risk mitigation is addressed. Then, we describe the supervised learning problem, based on a literature review covering the main concepts of this area, main applications and performance evaluation methods of the models used for classification tasks. Then, we do a literature review, describing some works in which classical and hybrid methods were used to detect fraudulent transactions. We also describe the main methodologies for balancing datasets that are applicable to the problem under analysis. At the end of the work, we bring the final considerations, also including some possibilities of studies for this area.