Research on Credit Card Fraud Detection Based on GAN

Abstract

With the advancement of technology, credit card transactions have become more and more popular, and the number of credit card frauds has also increased at the same time. In order to reduce property loss of cardholders and banks, many traditional machine learning algorithms based on binary classification are applied to detect fraudulent transactions. However, the number of fraudulent transactions and normal transactions in the dataset used to train the classifier is seriously imbalanced, which causes the classifier with the goal of accuracy to tend to classify all transactions as normal transactions. It is meaningless to detect fraudulent transactions with this trained classifier. Oversampling the minority fraudulent transaction classes to rebalance the training set is an effective way to address the problem of class imbalance. The method proposed in this paper uses an improved conditional generative adversarial network to rebalance dataset, then combines with random forests classifier for fraud detection, and experiments prove that the proposed method is effective for credit card fraud detection.