Scholarlink

Financial fraud detection is difficult owing to severe class imbalance and the fact that fraudulent activities can adapt. The paper introduces a hybrid model that incorporates Generative Adversarial Networks (GANs) to augment minority classes, the Synthetic Minority Oversampling Technique (SMOTE) to further balance, and Extreme Gradient Boosting (XGBoost) for classification. Experiments done on popular Kaggle credit card fraud data (284,807 transactions, 492 fraud instances) show that our method far outperforms conventional methods. The model proposed in this work has an F1-score of 0.7823 and an AUC of 0.9973, which indicates significant improvement in precision as well as recall. These results confirm the efficacy of integration of data-oriented and algorithmic methods for effective fraud detection and propose practical feasibility in real-world financial systems.

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

A Hybrid Approach for Financial Fraud Detection: GAN-Generated Synthetic Data and XGBoost with SMOTE Balancing

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Vivek S. Bharadwaj,
  • Dhanush Goudra,
  • Rajesh M Patil,
  • Aditya Billowria,
  • Sharadha K Shiragudikar

摘要

Financial fraud detection is difficult owing to severe class imbalance and the fact that fraudulent activities can adapt. The paper introduces a hybrid model that incorporates Generative Adversarial Networks (GANs) to augment minority classes, the Synthetic Minority Oversampling Technique (SMOTE) to further balance, and Extreme Gradient Boosting (XGBoost) for classification. Experiments done on popular Kaggle credit card fraud data (284,807 transactions, 492 fraud instances) show that our method far outperforms conventional methods. The model proposed in this work has an F1-score of 0.7823 and an AUC of 0.9973, which indicates significant improvement in precision as well as recall. These results confirm the efficacy of integration of data-oriented and algorithmic methods for effective fraud detection and propose practical feasibility in real-world financial systems.