Credit Card Fraud Detection: An Exploration of Different Sampling Methods to Solve the Class Imbalance Problem

Abstract

AbstractCredit card fraud is a plague that has been affecting the financial organizations and businesses since the time of its inception. Businesses incur loss of millions of dollars for such fraudulent transactions and at times customers also bear the brunt by losing their hard-earned money in disputes. Hence, the importance of detection of credit card fraud cannot be undermined. Credit card fraud is usually a problem of classification in the machine learning parlance, and the focus remains not only to identify fraud but also not to penalize the genuine customers, i.e., reduce the misclassification rate. The problem is rendered difficult because though the impact of the issue is paramount the occurrence is in-fact not that frequent. This poses the challenge of the detection of fraud through samples that are often imbalanced. This paper explores the various detection methods of predicting fraud correctly through data driven machine learning techniques, while focusing on the data imbalance problem and providing solutions on how to resolve it.KeywordsCredit card fraudClass imbalanceMachine learningMisclassificationUnder-samplingOver-samplingSMOTEDeployment