Causal Intent-Disentangled Feature Engineering for Fraud Detection in Government Expenditure
DOI:
https://doi.org/10.59890/ijgsr.v4i10.15Keywords:
Government Expenditure Fraud, Causal Representation Learning, Graph Neural Networks, Domain Generalization, Explainable AIAbstract
Government expenditure fraud models often confuse misconduct signals with benign administrative style, which weakens transfer across agencies, regions, procurement categories, and time. This study develops Causal Intent-Disentangled Feature Engineering (CIDFE), a heterogeneous-graph representation framework that separates fraud intent from contextual style using invariant prediction, adversarial domain confusion, orthogonality, counterfactual style swapping, and confounder-adjusted residuals. Leave-domain-out experiments show that CIDFE improves average and worst-domain AUPRC, reduces false positives for high-value legitimate contracts, and improves calibration relative to tabular, graph, textual, multimodal, and domain-generalization baselines. Shapley-based explanations further separate genuine red flags from administrative anomalies. The findings support CIDFE as a transferable, auditor-oriented feature engineering approach for public expenditure fraud detection
References
Deng, Z., Tian, H., Zheng, X., & Zeng, D. (2025). Deep causal learning: Representation, discovery and inference. ACM Computing Surveys.
Feng, L., Lin, A., & Luo, Z. (2026). MixingInsights: A framework for causal inference with confounder representation learning from mixed structured and textual data. Cognitive Computation.
Gara, M., Iezzi, S., & Siino, M. (2024). Corruption risk indicators in public procurement: A proposal using italian open data. Bahamasamlconference ….
Giudici, P., & Raffinetti, E. (2021). Shapley-lorenz eXplainable artificial intelligence. Expert Systems with Applications.
Gupta, D., Chugh, P., Sood, K., & Grima, S. (2026). … intelligence in financial fraud detection: A systematic review and FinTech-oriented ADO-TCCM meta-framework for trust, governance, and transparency. FinTech.
Li, B., Shen, Y., Wang, Y., Zhu, W., Li, D., Keutzer, K., et al. (2022). Invariant information bottleneck for domain generalization. Proceedings of the AAAI Conference on Artificial Intelligence.
Li, Y. (2024). Application of artificial intelligence in cross-departmental budget execution monitoring and deviation correction for enterprise management. Journal of Artificial Intelligence Review.
Man, H., & Nguyen, T. H. (2024). Counterfactual augmentation for robust authorship representation learning. Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval.
Odunaike, A. (2025). Designing adaptive compliance frameworks using time series fraud detection models for dynamic regulatory and risk management environments. Comput. Eng. Appl.
Salhan, S., Zhou, E., & Buttery, P. (2026). Do monolingual language models learn cross-lingual universal conceptual representations? Conference on Concept Representation Learning.
Sany, S. A., & Uddin, H. (2023). Machine learning-based fraud detection and conventional audit approaches in government deposit processing. American Journal of Interdisciplinary Studies.
Schiavon, J., Castro, M., & Carvalho, J. (2026). Ensemble-based anomaly detection in brazilian parliamentary expenses using unlabeled public data. International Journal of Data Science and Analytics.
Sha, Q., Tang, T., Du, X., Liu, J., Wang, Y., et al. (2025). Detecting credit card fraud via heterogeneous graph neural networks with graph attention. 2025 IEEE 6th International Seminar on Artificial Intelligence, Networking and Information Technology (AINIT).
Sharma, A., & Hassan, M. (2026). MissingMod-audit: Reliability-aware multimodal fraud detection under missing and corrupted financial evidence. International Journal of Science, Engineering and Information Technology.
Shovon, M. (2025). Application of machine learning and anomaly-detection algorithms for fraud detection in US government financial operations. American Journal of Interdisciplinary Studies.
Soares, H., Moura, R., Machado, V., Paiva, A., Lima, W., et al. (2025). The detection of spurious correlations in public bidding and contract descriptions using explainable artificial intelligence and unsupervised learning. Electronics.
Wang, J., Liu, J., Zheng, W., & Ge, Y. (2025). Temporal heterogeneous graph contrastive learning for fraud detection in credit card transactions. IEEE Access.
Yang, Y., Uy, M., & Huang, A. (2020). Finbert: A pretrained language model for financial communications. arXiv preprint arXiv:2006.08097.
Zhao, C., Wan, S., Sheth, P., Patwa, K., et al. (2026). Causality guided representation learning for cross-style hate speech detection. Proceedings of the ACM Web Conference.
Zhao, T., Liu, G., Wang, D., Yu, W., et al. (2022). Learning from counterfactual links for link prediction. International Conference on Machine Learning.
