ABSTRACT
Millions of transactions are registered daily in electronic banking systems. Most of them are legal operations, but a small percentage are attempts at illegal activities. An essential element of the fraud detection systems currently being developed is data mining, i.e., discovering significant and previously unknown patterns in large data sets. In this chapter, we look at the use of the generative adversarial networks (GAN) approach and is currently one of the most effective generative modeling. The use of neural networks allows one to obtain results unattainable with classical methods. However, they generally require much information to operate effectively and are also characterized by a prediction time that is often clearly longer than, for example, logistic regression.
