Leveraging generative AI for financial fraud detection in the platform economy: a GAN-based data augmentation method

As the platform economy develops, digital financial platforms have become central to economic transactions while facing increasing risks of financial fraud that threaten market integrity and user trust. Financial fraud not only hinders the development of the digital economy but also disrupts financial security. Existing research has focused extensively on the problem of financial fraud and proposed many algorithms to identify financial fraud accounts. However, the problems of data imbalance and data scarcity in financial fraud accounts detection have not been well addressed, which reduces the prediction precision of traditional machine learning models. Accordingly, we proposed a financial fraud detection method based on generative artificial intelligence technology, called GAN-FFAD, aiming to solve the problems of data imbalance and data scarcity in financial fraud accounts detection. Specifically, GAN is first trained separately using different data categories to generate vectors that match their feature distributions. Second, the feature vectors obtained from the generator of the GAN are merged into the feature vector matrix of the initial financial dataset to expand the training data. Results show that the GAN-FFAD method, as a generative artificial intelligence technology, could improve the performance of machine learning models in detecting financial fraud accounts.