Anomaly Detection in Credit Card Transactions

Abstract

We live a significant increase in the use of credit cards which leads to a high number of fraudulent transactions. The detection of fraudulent transactions carried out by the credit card is an important application in anomaly detection. The use of the credit card has not only several advantages but also losses and damages that may reach billions of dollars. However, existing approaches and methods are not optimized for detecting anomalies. When facing large volume of data, these methods remain limited resulting in a very high percentage of unsupported anomalies. The purpose of this paper is to compare these different techniques in order to choose the most adequate one to detect real-time anomalies in credit card transactions. We have opted for Isolation Forest which not only achieved a high-level of detection accuracy and AUC Score but also increased the rate of fraudulent transaction detection and minimized the percentage of incorrect fraud classifications. In this paper, we propose an anomaly detection model in order to predict and detect fraudulent transactions over time.