Exploratory Studies LAIA@UIB: AI models for the detection of fraudulent payments

Abstract

[eng] This thesis investigates the use of AI models for detecting fraudulent payments in electronic payment systems. The main challenges in the development of such models are the lack of labeled data, the need to balance minimizing false positives while maximizing true positives, and the complexity of financial transactions. This study aims to explore the performance of different machine learning and deep learning algorithms, such as logistic regression, neural networks, XGBoost, and Random Forest, in detecting fraudulent payments, and to develop techniques to address data scarcity and imbalance. The research involves experimentation with two datasets, one real and the other artificially generated, both exhibiting a high degree of imbalance. The study findings can enhance the development of trustworthy and effective AI models for the detection of fraudulent payments, contributing to enhancing security measures within financial systems.