An Improved Variational Autoencoder Generative Adversarial Network with Convolutional Neural Network for Fraud Financial Transaction Detection

In recent years, the rapid development of ecommerce technologies has made it possible for people to conveniently shop from stores worldwide without leaving their homes. Unfortunately, credit card fraud has become common due to online payments. This fraudulent activity causes significant financial losses, and financial institutions need to install automatic deterrent mechanisms to check these actions. Fraudulent transactions do not follow a specific pattern and continuously change their shape and behavior, making it difficult to detect them. To overcome these problems this paper proposed an improved generator part of the Variational Autoencoder Generative Adversarial Network (VAEGAN) along and introduces a new oversampling method that generates convincing and diverse minority class data. Then in the classification process Convolutional Neural Network (CNN) uses the parameters for the classification process. This enhances fraud detection in the transaction and improves the detection accuracy. The performance of the proposed model is measured in terms of accuracy, precision, recall, and F1-score. This shows that the proposed method outperforms existing methods such as CNN, LightGBM, and Long Short-Term Memory (LSTM) ensemble in terms of accuracy of 99.78%, precision of 88.97%, recall of 95.24, and f1-score of 95.00%.