Enhancing Fraud Detection in Financial Transactions Using Generative Adversarial Networks

Author(s): Adarsh Naidu

Publication #: 2503095

Date of Publication: 07.04.2019

Country: United States

Pages: 1-6

Published In: Volume 5 Issue 2 April-2019

DOI: https://doi.org/10.5281/zenodo.15087334

Abstract

Fraudulent activities in financial transactions pose a major threat to the stability of financial systems, leading to significant economic losses and diminishing customer trust. Conventional fraud detection methods frequently struggle with the challenge of imbalanced datasets, where legitimate transactions far outnumber fraudulent ones, leading to suboptimal model performance and elevated false positive rates. This study examines the potential of Generative Adversarial Networks (GANs) in generating synthetic fraudulent transaction data, thereby enhancing the dataset used for fraud detection systems. By employing GANs, we synthesize fraudulent transactions that closely resemble real-world fraudulent activities, improving the accuracy of machine learning models in detecting fraud. Our experimental results indicate a substantial increase in fraud detection rates while simultaneously reducing false positives, demonstrating the transformative potential of GANs in financial fraud detection. This technique not only strengthens fraud detection mechanisms but also ensures adaptability to evolving fraud patterns, providing a scalable and effective solution for the financial sector (Goodfellow et al., 2014; Mirza & Osindero, 2014; Salimans et al., 2016).

Keywords: Fraud Detection, Financial Transactions, Generative Adversarial Networks, Imbalanced Datasets, Synthetic Data, Machine Learning, False Positives, Precision, Recall, Adaptability.

Download/View Paper's PDF

Download/View Count: 94

Share this Article