Applying Explainable Graph Neural Networks for Credit Risk Prediction and Fraud Detection

Mahmood, Mehrab (2025) Applying Explainable Graph Neural Networks for Credit Risk Prediction and Fraud Detection. Masters thesis, Dublin, National College of Ireland.

Preview

PDF (Master of Science) Download (490kB) | Preview

Abstract

Financial institutions are increasingly reaching out to machine learning to enable them in making important decisions like credit risk analysis and fraud detection. nevertheless, standard machine learning frameworks are not always able to encapsulate complicated connections among customers, accounts, and transactions.

This research, hypothesizes and compares explainable Graph Neural Network (GNN) designs in credit risk prediction and credit fraud detection. there are two popular GNN architectures, namely Graph Convolutional Networks (GCNs) and Graph Attention Networks (GATs), which are provided through Pytorch Geometric. in order to enhance model transparency an extended version of GNN-explainer is used to produce structural explanations that indicate the influential subgraphs whereas SHAP is utilized to determine which input features play the biggest role in predictions.

The models are tested and trained using publicly available data and the data includes the home credit default risk dataset and the IEEE-CIS credit card Fraud dataset.

The results indicate that the proposed strategy may be useful in the identification of high-risk customers, the identification of suspicious transaction patterns, and also identifying important factors determining model decisions.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Simiscuka, Anderson 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 H Social Sciences > HG Finance > Credit. Debt. Loans. Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Artificial Intelligence
Depositing User:	Ciara O'Brien
Date Deposited:	02 Sep 2026 10:20
Last Modified:	02 Sep 2026 10:20
URI:	https://norma.ncirl.ie/id/eprint/9763

Actions (login required)

View Item