Estudo comparativo entre modelos para detecção de fraudes em cartões de crédito

Abstract

Fraud, as a criminal act, has become commonplace when it comes to online credit card purchases. In this context, the work of researchers and students through machine learning models that solve the problem has become well targeted, since this problem plagues financial intuitions and companies that provide cards. Therefore, this work aims to carry out a comparative study between three models of machine learning and to observe which of them react better to the detection with specific types of fraudulent acts obtained in the training. First, an assessment was carried out in the field of fraud involving cards and purchases, then a treatment on the database (DataSet) to then apply the machine methods. Respectively: Decision Trees, Naive Bayes Classification and Support Vector Machine. Subsequently, demonstrated in evaluation by graphics, showing the percentage of learning based on types of fraud in the detection made by the machine.