CCFD-Net: A novel deep learning model for credit card fraud detection

Xiao Liu, Kuan Yan, Levent Burak Kara, Zhenguo Nie

Abstract: Credit card fraud can cause billions of dollars in financial losses to merchants and consumers each year. If fraud can be detected in time and corresponding measures can be taken, financial losses can be significantly alleviated and other derivative frauds can be prevented. Although traditional machine learning methods can achieve good precision and recall in credit card fraud detection, they cannot avoid the false positives effectively. In this paper, we propose a novel Credit Card Fraud Detection model called CCFD-Net that employs a modified residual network architecture. Based on a realworld dataset from Vesta's e-commerce transactions, we conduct comparative analysis on predictive models to evaluate and verify the effectiveness of the proposed method. The paper explores a hybrid architecture of 1D-Conv and the residual neural network (Res-net), evaluates the performance of different machine learning models based on K-fold cross-validation. The results prove the effectiveness and robustness of the model in credit card fraud detection. In practice, our proposed model can identify more fraudulent transactions than other compared models, and performs best on the evaluation metrics. We publicly share our full implementation with the dataset and trained models at https://github.com/zhenguonie/2021_CCFD_Net.