Dense Neural Networks and Convolutional Neural Networks for Fraud Detection in Transactional Movements

e451

Abstract

Currently, banking fraud affects both businesses and users of financial institutions, leading to significant economic losses that undermine trust in payment systems and electronic transactions. Traditional methods for detecting fraudulent activities, which rely on predefined rules and manual analysis, are insufficient to handle the increasing volume and complexity of data. This study applies Deep Learning (DL) techniques and models, such as Dense Neural Networks (DNN) and Convolutional Neural Networks (CNN), with the objective of detecting fraudulent transactions in the financial sector, thereby contributing to the security of online banking systems. The combined implementation of DNN and CNN models produced positive results, achieving an AUC-ROC greater than 0.8 and a sensitivity exceeding 80%. These results indicate that the models can detect a significant number of fraudulent activities without compromising overall system accuracy. Therefore, the implementation of deep neural networks, along with their integration with different architectures, represents a promising approach for fraud detection in the banking system.