Credit Card Fraud Detection: Comparison of Different Machine Learning Techniques

Abstract

Credit Card Fraud can be defined as the situation in which a person uses someone else's credit card for personal reasons and the cardholder and the card issuer are unaware of the use of the relevant credit card. In this study, credit card fraud detection models were developed. R programming language was used. During the engineering applications of this model, different machine learning algorithms were used for the same data set. Relevant performance curves were drawn for the models. Data has been analyzed and visualized to distinguish credit card fraud from other data types. In this context, different machine learning techniques used. These are logistic regression, decision tree, neural networks and gradient boosting. The aforementioned techniques were used to solve the problem of fraud/not fraud within the scope of the study. However, the results obtained are generally give idea for all classification problems. According the findings of this study, the fastest results among these techniques were found with artificial neural networks. Gradient boosting and decision tree technique are related to each other. It is a collection of weak prediction models with gradient boosting decision trees. The most effective and well-known technique from these four techniques is logistic regression.