Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

Uncertainty-Aware Quantum Financial Intelligence: Hybrid Transformer-Graph Architectures for Fraud Detection and Transaction Screening

GODWIN OLUWAFEMI OLAOYE

Published Jun 16, 2026• 0 citations
Publisher
DOI
10.5281/zenodo.20718432

Related papers

Quantum-Enhanced Financial Crime Detection with Hybrid Transformers, Entity Resolution, and Risk-Aware Evidence Graph Discovery

GODWIN OLUWAFEMI OLAOYE

•Jun 16, 2026•0 citations

Financial fraud detection using quantum graph neural networks

Nouhaila Innan, Abhishek Sawaika, Ashim Dhor +7 more

arXiv:2309.01127•Quantum Machine Intelligence•Feb 1, 2024•149 citations

Hybrid Quantum Graph Transformers with Conformal Risk Guarantees for Explainable Fraud Detection and Anti-Money Laundering

GODWIN OLUWAFEMI OLAOYE

•Jun 17, 2026•0 citations

Hybrid Quantum-Classical Fraud Detection Using Gated Token Mixing Transformers and Graph-Based Evidence Subgraph Discovery for Explainable AML

GODWIN OLUWAFEMI OLAOYE

•Jun 16, 2026•0 citations

Quantum Chebyshev transform-based graph neural networks for financial fraud detection

Minrui XU, Bingyan Guan, Bethel Hui Ting LOKE +1 more

Financial fraud detection is a critical challenge requiring accurate identification of anomalous patterns in complex transaction networks. Graph Neural Networks (GNNs) have emerged as powerful tools for fraud detection by capturing relational structures among entities. Meanwhile, quantum computing offers new possibilities to enhance machine learning through high-dimensional Hilbert spaces and parallelism. In this paper, we propose a hybrid classical-quantum model called QCTGNN (Quantum Chebyshev Transform-based Graph Neural Network) for financial fraud detection. The QCTGNN integrates a classical graph neural network component based on Simplified Graph Convolutions (SGConv) with a quantum component that performs a Chebyshev polynomial-based transform via variational quantum circuits. The quantum component employs angle encoding of node features, Chebyshev polynomial embedding to enrich feature representations, multiple layers of parameterized RY rotations with entangling gates, and Pauli-Z measurements for readout. To address the noise and instability of Noisy Intermediate-Scale Quantum (NISQ) devices, we incorporate Noise-Aware Quantum Calibration (NQC) that adjusts for decoherence and readout bias in the quantum outputs. Our system also supports adaptive graph construction using correlation thresholds and temporal windows to capture dynamic transaction patterns, and uses SMOTE oversampling to rebalance class distributions in training data only. We present the architecture and algorithmic design of QCTGNN and provide an experimental evaluation on four financial transaction datasets. Across these benchmarks, QCTGNN matches or exceeds classical GNN baselines in ROC-AUC and PR-AUC, demonstrating the potential of quantum-enhanced graph learning for financial fraud analytics.

•Feb 1, 2026•0 citations

Design of an Iterative Explainable Quantum Machine Learning Method for Adaptive Fraud Detection in Financial Transaction Systems

Pramoda Patro Mudimela Madhusudhan

Financial transaction systems are becoming increasingly complex and large, in this scenario, fraud detection patterns have become more sophisticated than the conventional detection mechanisms by themselves. While machine learning models have been developed for operational purposes, very few of them are interpretable, sensitive to class imbalance or do not generalize to new fraudulent behaviors. Research on quantum machine learning (QML) for fraud detection is still nascent, with many efforts concentrating on classification without addressing explainability or even attempting to deal with graph-structured and sequential data comprehensively. This study aims to fill this gap by presenting a new explainable quantum machine learning architecture to detect fraud from financial transaction datasets. Three core quantum models are developed and validated: (1) Hybrid Quantum-Loss Adaptive Variational Autoencoder (HQLAVAE) for anomaly detection-has a fidelity aware, class-adaptive loss function that enhances reconstruction of typical transaction patterns and sensitively detects anomalies; (2) The third is Quantum Feature Relevance Graph Attribution (Q-FRGA): an explainable quantum graph neural network particularly capturing transactional relationships with respect to node-level fraud-attributed predictions from Hilbert-space-based relevance scoring; and (3) Quantum Environment-Adaptive Policy Refinement (Q-EAPR), a reinforcement learning model integrating distributional quantum Bellman updates with a novel tracking mechanism for policies to enable online processing of fraud behavior changes. Moreover, two cross-model innovations for enhanced explainability and model integration include Cross-Modal Quantum Counterfactual Explainer (CQCE), generating explainable perturbation-based rationales for single predictions as well as Quantum Residual Explainability-Integrated Ensemble Framework (Q-REIEF) using an ensemble of all models with residual error corrections and meta-explanation tracing sets. This work further advances quantum explainable AI in the areas of transparent, adaptive and high-accuracy fraud detection-all critical gaps in interpretability, data imbalance and behavioral drift on financial systems.

•Jun 1, 2026•0 citations
Financial fraud detection is a critical challenge requiring accurate identification of anomalous patterns in complex transaction networks. Graph Neural Networks (GNNs) have emerged as powerful tools for fraud detection by capturing relational structures among entities. Meanwhile, quantum computing offers new possibilities to enhance machine learning through high-dimensional Hilbert spaces and parallelism. In this paper, we propose a hybrid classical-quantum model called QCTGNN (Quantum Chebyshev Transform-based Graph Neural Network) for financial fraud detection. The QCTGNN integrates a classical graph neural network component based on Simplified Graph Convolutions (SGConv) with a quantum component that performs a Chebyshev polynomial-based transform via variational quantum circuits. The quantum component employs angle encoding of node features, Chebyshev polynomial embedding to enrich feature representations, multiple layers of parameterized RY rotations with entangling gates, and Pauli-Z measurements for readout. To address the noise and instability of Noisy Intermediate-Scale Quantum (NISQ) devices, we incorporate Noise-Aware Quantum Calibration (NQC) that adjusts for decoherence and readout bias in the quantum outputs. Our system also supports adaptive graph construction using correlation thresholds and temporal windows to capture dynamic transaction patterns, and uses SMOTE oversampling to rebalance class distributions in training data only. We present the architecture and algorithmic design of QCTGNN and provide an experimental evaluation on four financial transaction datasets. Across these benchmarks, QCTGNN matches or exceeds classical GNN baselines in ROC-AUC and PR-AUC, demonstrating the potential of quantum-enhanced graph learning for financial fraud analytics.
Financial transaction systems are becoming increasingly complex and large, in this scenario, fraud detection patterns have become more sophisticated than the conventional detection mechanisms by themselves. While machine learning models have been developed for operational purposes, very few of them are interpretable, sensitive to class imbalance or do not generalize to new fraudulent behaviors. Research on quantum machine learning (QML) for fraud detection is still nascent, with many efforts concentrating on classification without addressing explainability or even attempting to deal with graph-structured and sequential data comprehensively. This study aims to fill this gap by presenting a new explainable quantum machine learning architecture to detect fraud from financial transaction datasets. Three core quantum models are developed and validated: (1) Hybrid Quantum-Loss Adaptive Variational Autoencoder (HQLAVAE) for anomaly detection-has a fidelity aware, class-adaptive loss function that enhances reconstruction of typical transaction patterns and sensitively detects anomalies; (2) The third is Quantum Feature Relevance Graph Attribution (Q-FRGA): an explainable quantum graph neural network particularly capturing transactional relationships with respect to node-level fraud-attributed predictions from Hilbert-space-based relevance scoring; and (3) Quantum Environment-Adaptive Policy Refinement (Q-EAPR), a reinforcement learning model integrating distributional quantum Bellman updates with a novel tracking mechanism for policies to enable online processing of fraud behavior changes. Moreover, two cross-model innovations for enhanced explainability and model integration include Cross-Modal Quantum Counterfactual Explainer (CQCE), generating explainable perturbation-based rationales for single predictions as well as Quantum Residual Explainability-Integrated Ensemble Framework (Q-REIEF) using an ensemble of all models with residual error corrections and meta-explanation tracing sets. This work further advances quantum explainable AI in the areas of transparent, adaptive and high-accuracy fraud detection-all critical gaps in interpretability, data imbalance and behavioral drift on financial systems.