Recognizing Credit Card Fraud Using Machine Learning Methods

Abstract

The rapid growth of e-commerce and online-based payment possibilities carries along with an empirical universe of economic fraud in which credit card fraud is more preventing for many years; several researchers have developed many data mining base methods to overcome this problem. Recently there has been a major interest in applying machine learning algorithms in place of data mining techniques to discover credit card fraud. Continuous work is done to bring in a conceptual difference between fraud recognition and forecasting probable non genuine opportunities in the digital space of financial transactions. This research shows several algorithms that can be used to solve credit card fraud, however, a number of challenges appear, such as dataset imbalance, variant fraudulent behaviour, etc. In this paper, we are going to address the problem of an imbalanced dataset. SMOTE sampling technique is used to convert the imbalanced dataset to a balanced binary dataset. There are now various machine learning algorithms that are tested by using the European credit card dataset and are compared by using evaluation metrics like accuracy, precession, AUC value, and ROC curve, etc. The results are very encouraging and graphs have been plotted which identify the best suitable algorithm.