Anomaly Detection using Generative Adversarial Network VAE

Abstract

Fraudulent credit card transactions continue to be one of the problems facing businesses and banks. It causes us to lose billions of dollars each year. Designing efficient algorithms is one of the most important challenges in this field. This paper aims to propose an efficient approach that automatically detects fraud in credit card transactions using Generative Adversarial Network Variational Auto encoders. The effectiveness of the proposed method (Generative Adversarial Network VAE) has been proved in identifying fraud in actual data from transactions made by credit cards. However, the typical credit card data set presents an imbalanced classification landscape due to highly skewed class distributions. Researchers have proposed several strategies to address these imbalances, but draw backs still remain. The proposed method is tested on an open credit card fraud dataset, which contains 20 million transactions generated from a multi-agent virtual world simulation performed by IBM. Experimental results show that the VAE method performs better than traditional deep neural network methods. Through experiments, compared with the VAE and traditional fully connected neural networks, the results showed the proposed algorithm improves the classification accuracy of a minority class of imbalanced datasets.