Developing Real-Time Fraud Detection and Response Mechanisms for Financial Transactions

• Author(s): Adeyinka Orelaja ; Adenike F. Adeyemi
• Paper ID: 1706044
• Page: 573-582
• Published Date: 06-08-2024
• Published In: Iconic Research And Engineering Journals
• Publisher: IRE Journals
• e-ISSN: 2456-8880
• Volume/Issue: Volume 8 Issue 1 July-2024

Abstract

The emphasis on real-time fraud detection and response within the financial sector is not only motivated by the growing threat of fraudulent schemes but by regulatory demands and the importance of trust between the consumers and the organizations. This study employed the IEEE-CIS Fraud Detection dataset, which contains attributes pertaining to transaction and client identities, together with labels indicating the fraudulent or non-fraudulent nature of the transactions. To adequately capture the relationships and interactions inside the transaction network, a Graph Neural Network (GNN) model was built due to the intricate and ever-changing nature of fraud patterns. The GNN utilizes the inherent organization of the data, hence improving its capacity to detect fraudulent actions. The findings of this study showed the model's accuracy, precision, and recall as 0.9981, 0.9981, and 0.866 respectively. The 99.81% precision attained by the model signifies its ability to accurately forecast the bulk of transactions. Nevertheless, relying just on accuracy can be deceptive when dealing with imbalanced datasets, characterized by a significantly lower number of fraudulent transactions compared to valid ones. Minimizing false positives is also vital in fraud detection as it helps to reduce unneeded investigations or inconveniences for customers, thus, the recall detection rate of 86.6% signifies that the model accurately detects 86.6% of all fraudulent transactions. This study recommends further research in enhancing recall to minimize the number of fraudulent transactions that remain unnoticed. It also suggests the integration of explainable artificial intelligence (XAI) to enhance comprehensibility of models embedded into Graph Neural Networks.

Keywords

Fraud Detection, Fraudulent Transactions, Graph Neural Networks, Recall

Citations

IRE Journals:
Adeyinka Orelaja , Adenike F. Adeyemi "Developing Real-Time Fraud Detection and Response Mechanisms for Financial Transactions" Iconic Research And Engineering Journals Volume 8 Issue 1 2024 Page 573-582

IEEE:
Adeyinka Orelaja , Adenike F. Adeyemi "Developing Real-Time Fraud Detection and Response Mechanisms for Financial Transactions" Iconic Research And Engineering Journals, 8(1)