Scholarlink

Credit card fraud detection is vital for mitigating financial losses, yet the scarcity of fraudulent transactions results in imbalanced datasets, posing challenges for model training. This study employs Generative Adversarial Networks (GANs) to generate synthetic fraud cases, achieving a balanced dataset. We investigate the performance of four deep learning architectures—Deep Neural Network (DNN), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU)—in identifying fraudulent transactions. These models are evaluated using a robust set of metrics: accuracy, precision, recall, F1-score, Area Under the Curve (AUC), and Matthews Correlation Coefficient (MCC). The findings highlight the strengths of each architecture, offering valuable insights into their effectiveness for fraud detection. By integrating GAN-based data augmentation with deep learning, this study enhances the precision and reliability of credit card fraud detection systems, contributing significantly to the field.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

GAN-Augmented Deep Learning for Robust Credit Card Fraud Detection

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Mohamed Akouhar,
  • Mohamed El Fatini,
  • Abdallah Abarda,
  • Mohamed Ouhssini,
  • El hafed Agherrabi

摘要

Credit card fraud detection is vital for mitigating financial losses, yet the scarcity of fraudulent transactions results in imbalanced datasets, posing challenges for model training. This study employs Generative Adversarial Networks (GANs) to generate synthetic fraud cases, achieving a balanced dataset. We investigate the performance of four deep learning architectures—Deep Neural Network (DNN), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU)—in identifying fraudulent transactions. These models are evaluated using a robust set of metrics: accuracy, precision, recall, F1-score, Area Under the Curve (AUC), and Matthews Correlation Coefficient (MCC). The findings highlight the strengths of each architecture, offering valuable insights into their effectiveness for fraud detection. By integrating GAN-based data augmentation with deep learning, this study enhances the precision and reliability of credit card fraud detection systems, contributing significantly to the field.