Causal Temporal Graph Neural Networks with Cost-Aware Reinforcement Learning for Adaptive Financial Fraud Detection
DOI:
https://doi.org/10.66421/int.j.datasci.intell.syst74Keywords:
Financial fraud, temporal graph neural networks, reinforcement learning, causal evaluation, concept drift, decision costAbstract
Financial fraud screening couples relational inference with a sequential operational decision: which transactions should be escalated under asymmetric investigation and missed-fraud costs. Building on the graph-to-decision design of FraudGNN-RL, which couples relational risk scoring with reinforcement-learning threshold control, we develop a causal temporal transaction graph and a cost-aware discrete Q-learning threshold controller with validation-based safety gating. Directed predecessor edges are created only from earlier transactions sharing a sender or merchant, so that a trainable message-passing detector uses historical messages without accessing later transactions, and the controller selects a threshold for each transaction window from information available in preceding labeled windows. Experiments on three independently seeded simulated streams of 12,000 transactions with a fraud-regime shift use strict chronological train/validation/test partitions (60/20/20) and an eight-to-one false-negative versus false-positive cost. The safety gate keeps held-out decision cost at the level of the best validation-tuned threshold (0.273 per transaction on average) in every stream, whereas an ungated Q policy raises it to 0.403, showing that validation-based gating is essential for deploying adaptive thresholds under drift. The study contributes a leakage-free causal graph specification, a cost-aware adaptive thresholding scheme with safety gating, and a reproducible benchmark for identifying when graph enrichment and reinforcement-based threshold adaptation are beneficial.References
F. Zhou, “Moment-constrained side-force allocation for wind-assisted merchant ships: Hydrodynamic performance of retractable anti-leeway fins and flap rudders,” Innovation and Technology Studies, vol. 3, no. 1, pp. 27–41, 2026.
J. West and M. Bhattacharya, “Intelligent financial fraud detection: A comprehensive review,” Computers & Security, vol. 57, pp. 47–66, 2016.
A. Dal Pozzolo, G. Boracchi, O. Caelen, C. Alippi, and G. Bontempi, “Credit card fraud detection: A realistic modeling and a novel learning strategy,” IEEE Trans. Neural Networks and Learning Systems, vol. 29, no. 8, pp. 3784–3797, 2018.
J. Gama, I. Žliobaitė, A. Bifet, M. Pechenizkiy, and A. Bouchachia, “A survey on concept drift adaptation,” ACM Computing Surveys, vol. 46, no. 4, art. 44, 2014.
B. Lebichot, Y.-A. Braun, O. Caelen, and G. Bontempi, “A taxonomy of supervised learning for concept drift in credit card fraud detection,” ACM Computing Surveys, vol. 51, no. 6, 2019.
D. Teng, “RSRBench: Reproducibility and performance benchmarking for regional seismic risk workflows,” Academic Journal of Architecture and Civil Engineering, vol. 3, no. 1, pp. 40–52, 2025.
T. N. Kipf and M. Welling, “Semi-supervised classification with graph convolutional networks,” Proc. ICLR, 2017.
W. L. Hamilton, R. Ying, and J. Leskovec, “Inductive representation learning on large graphs,” Proc. NeurIPS, pp. 1024–1034, 2017.
J. Wang, Y. Ge, J. Liu, and B. Jiang, “Early warning of liquidity stress in U.S. banks with seasonally aligned frailty hazards and delayed-feedback conformal alarms,” Journal of Trends in Finance and Economics, vol. 1, no. 1, pp. 85–100, 2024.
Y. Dou, Z. Liu, L. Sun, Y. Deng, H. Peng, and P. S. Yu, “Enhancing graph neural network-based fraud detectors against camouflaged fraudsters,” Proc. CIKM, pp. 315–324, 2020.
Z. Liu, Y. Dou, P. S. Yu, Y. Deng, and H. Peng, “Alleviating the inconsistency problem of applying graph neural network to fraud detection,” Proc. SIGIR, pp. 1569–1573, 2021.
Y. Cui, X. Han, J. Chen, X. Zhang, J. Yang, and X. Zhang, “FraudGNN-RL: A graph neural network with reinforcement learning for adaptive financial fraud detection,” IEEE Open Journal of the Computer Society, vol. 6, pp. 426–437, 2025, doi: 10.1109/OJCS.2025.3543450.
D. Xu, C. Ruan, E. Korpeoglu, S. Kumar, and K. Achan, “Inductive representation learning on temporal graphs,” Proc. ICLR, 2020.
E. Rossi, B. Chamberlain, F. Frasca, D. Eynard, F. Monti, and M. Bronstein, “Temporal graph networks for deep learning on dynamic graphs,” arXiv:2006.10637, 2020.
M. Weber et al., “Anti-money laundering in Bitcoin: Experimenting with graph convolutional networks for financial forensics,” arXiv:1908.02591, 2019.
J. Wang, J. Liu, Y. Ge, and B. Jiang, “Risk-aware ensemble learning for credit default prediction under economic uncertainty,” World Journal of Economics and Business Research, vol. 1, no. 1, pp. 18–32, 2023.
V. Mnih et al., “Human-level control through deep reinforcement learning,” Nature, vol. 518, pp. 529–533, 2015.
H. van Hasselt, A. Guez, and D. Silver, “Deep reinforcement learning with double Q-learning,” Proc. AAAI, pp. 2094–2100, 2016.
C. Whitrow, D. J. Hand, P. Juszczak, D. Weston, and N. M. Adams, “Transaction aggregation as a strategy for credit card fraud detection,” Data Mining and Knowledge Discovery, vol. 18, pp. 30–55, 2009.
X. Yue and R. Zhu, “An explainable hybrid machine learning framework for cash-flow-conditioned multi-horizon liquidity risk early warning: An SME-oriented benchmark study,” World Journal of Management Science, vol. 2, no. 1, pp. 63–72, 2024.
H. Fan, Y. Jiao, M. Wang, J. Chen, and S. Yang, “Fraud learns too: continual graph learning under strategic adversarial drift in dynamic networks,” Scientific Reports, article in press, 2026, doi: 10.1038/s41598-026-60997-7.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Andre Henderson, Irene Ghosh, Lena Qian

This work is licensed under a Creative Commons Attribution 4.0 International License.
All articles are published under the Creative Commons Attribution 4.0 (CC BY 4.0) License. Users may share, copy, adapt, and distribute the work for any purpose, provided proper credit is given to the original author(s) and source. Authors retain copyright.
