Skip to content
Conference Open access

Evolutionary Multi-Objective Graph Neural Network for Adaptive Financial Fraud Detection in Dynamic Transaction Networks

Sep 2026 · 2026 10th International Conference on I-SMAC (IoT in Social, Mobile, Analytics and Cloud) (I-SMAC) · pp. 1861-1868 · 0 citations · 20 references

Abstract

Financial fraud detection is still complex and difficult because the financial transaction systems have created large scale, imbalanced, heterogeneous and continuous evolving transaction networks and fraudulent behaviours are fast changing. Current graph-based and deep-learning solutions enhance relational representation, but offer limited capabilities in terms of jointly optimizing the detection accuracy, false positive control, computational efficiency, and adaptability to new fraud patterns. This study introduces an Adaptive Financial Fraud Detection System Based on Evolutionary Multi-Objective Graph Neural Network (EMO-GNN) in a dynamic transaction network. The framework builds temporal transaction graphs, extracts graph features (both structural and behavioral), uses multi-objective evolutionary search to optimize a graph neural network to maximize recall of fraud transactions, maximize F1-score, minimize false-positive rate and computational cost. Experimental results show that EMO-GNN exceeds the performance of the conventional machine-learning, ensemble, GNN and dynamic graph-learning methods by achieving 99.21% accuracy, 96.84% precision, 95.73% recall, 96.28% F1-score and 98.71% AUC, and reduces the false-positive rate by 27.46% and detection latency by 21.38% against the best baseline. The framework also ensures strong performance in the face of changing transaction patterns. The results herein prove that evolutionary multi-objective optimization is capable of improving adaptive graph learning, and can offer an accurate, efficient and scalable solution to dynamic financial fraud detection.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.