Machine Learning and Credit Card Fraud

a systematic review

Abstract

Artificial Intelligence (AI) has become increasingly present in human life, pointing to an integrated future with the real and the virtual walking side by side. Inserted as an AI field, Machine Leaning operates with a set of tools, called algorithms, that allow performing activities in almost every area of knowledge, whether in Biology, Math and Science or Humanities. An application of the algorithms is in the area of credit card fraud detection, which causes financial institutions and society to lose billions of dollars every year. In this sense, this work aimed to identify how Machine Learning has contributed to this financial and social problem. For this proposal, it was used the Systematic Bibliographic Review, from which it was obtained primary diagnoses on the areas of Machine Learning, Fraud and Credit Card Fraud, in order to arrive at a thorough portrait of the subject. As a source of research, it was used in the Web of Science portal, with searches centered on the period from 2008 to 2018. The results show that, both in the subject of Machine Learning and fraud, most of the works refer to areas of Medicine & Biomedicine. In the integration between “Machine Learning” and credit card fraud, five papers were found that, when fully evaluated, proved to have good fraud detection power. There is, however, a way to go that allows comparing the methods employed from a common database.