AI-Powered Anomaly Detection for Fraudulent Credit Card Transactions

Devlekar, Om Sandeep (2025) AI-Powered Anomaly Detection for Fraudulent Credit Card Transactions. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (1MB) | Preview

Abstract

The proliferation of digital transactions has led to a parallel and alarming rise in credit card fraud, posing significant financial and reputational risks to consumers and institutions. Traditional fraud detection systems, often reliant on static rules, are increasingly insufficient against the sophisticated and evolving tactics of fraudsters. This research addresses the critical need for more dynamic and intelligent fraud detection mechanisms by leveraging advanced artificial intelligence (AI). This project provides a detailed, contrastive analysis of three different AI-based anomaly detection methods on the intrinsically difficult IEEE-CIS Fraud Detection dataset. The anomaly detection methods included an unsupervised entry (Isolation Forest), a semi-supervised model based on deep learning (Autoencoder), and a supervised deep learning based model (Long Short Term Memory network, LSTM). Using a strict pipeline for data processing, exploratory data analysis, feature engineering and extraction, and dimensional reduction with Principal Component Analysis (PCA) was strictly adhered to. Anomalous events from the model's perspective were evaluated based on the high class imbalance scenario in the data, and by comparing F1-Score, Precision and Recall. In comparison of the three methods, our findings found LSTM as the clear leader (with an F1-Score of 0.5312), followed by Isolation Forest (0.2315) and Autoencoder (0.2604). The overall success by LSTM was largely predicated upon it's high precision, and therefore suggested the model has a defined level of confidence in identifying true fraud. Our project eventually concluded with a proof of concept web application based on the most successful model. Overall, the results provide insights into the extent we might utilize supervised deep learning-based algorithms to perhaps improve the performance and reliability of existing tools for high finance security.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Rifai, Hicham UNSPECIFIED
Subjects:	H Social Sciences > HG Finance Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence Q Science > QA Mathematics > Computer software > Computer Security T Technology > T Technology (General) > Information Technology > Computer software > Computer Security H Social Sciences > HG Finance > Credit. Debt. Loans.
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	01 Jul 2026 10:00
Last Modified:	01 Jul 2026 10:00
URI:	https://norma.ncirl.ie/id/eprint/9424

Actions (login required)

View Item