Deontic-Temporal Policy-Verified Adaptive Ensembles with Safe Reinforcement Learning for AI-Driven Real-Time Tax Compliance Monitoring
DOI:
https://doi.org/10.24256/kharaj.v7i4.12929Keywords:
tax compliance monitoring; anomaly detection; policy verification; safe reinforcement learning; explainable AIAbstract
Tax compliance monitoring increasingly depends on timely detection of anomalous transactions, yet a prediction alone is insufficient when an administrative decision must also be justified against applicable rules. This study develops a compliance-aware anomaly detection architecture for an Integrated Tax Management Information System (ITMIS). It combines a heterogeneous ensemble of isolation forest, LSTM autoencoder, and gradient-boosted tree models with conformal uncertainty estimates, deontic-temporal policy monitoring, a constrained reinforcement-learning threshold governor, and cryptographically traceable decision records. The source manuscript reports an AUPRC of 0.847 on a synthetic deontic tax transaction stream, compared with 0.712 and 0.668 for two adaptive baselines. It also reports a demographic parity difference of 0.021, a gate rejection rate of 6.8%, and mean end-to-end latency of 47 ms. These figures are presented as reported study outcomes rather than independently reproduced measurements. The proposed contribution is to connect policy verification to the decision to accept or reject an adaptive threshold update, rather than reserving compliance checks for post hoc review. The framework offers a design for combining anomaly sensitivity, interpretable decisions, fairness monitoring, and tamper-evident auditing. Its operational and legal guarantees remain conditional on the correctness of encoded rules, monitoring semantics, and the underlying implementation.
References
Angelopoulos, A., & Bates, S. (2023). Conformal prediction: A gentle introduction. Foundations and Trends in Machine Learning.
Barik, M. (2026). Designing and implementing AI-based systems for effective corporate tax planning and management. capdr.org.
Cerqueira, V., Gomes, H., Heyden, M., et al. (2026). A framework for evaluating and benchmarking concept drift detection methods. Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining v.2.
Chen, T., & Guestrin, C. (2016). Xgboost: A scalable tree boosting system. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining.
Dudu, O., Alao, O., & Alonge, E. (2024). Conceptual framework for AI-driven tax compliance in fintech ecosystems. Journal of Fintech and Taxation.
Ha, V. (2026). Explainable artificial intelligence for anomaly detection in accounting and financial data: A machine learning and SHAP-based framework. Journal of Intelligent Decision Making and Information Sciences.
Huang, Y. (2024). Fairness-aware credit risk assessment using alternative data: An explainable AI approach for bias detection and mitigation. Journal of Artificial Intelligence Review.
Iqbal, M., Ali, M., Hina, U., et al. (2025). The future of smart tax systems: Integrating artificial intelligence, blockchain, and autonomous compliance technologies for transparent and efficient tax administration. Journal of Social Sciences and Economics.
Jamroga, W. (2015). Logical methods for specification and verification of multi-agent systems. webspace.science.uu.nl.
Li, Y., & Ling, Z. (2026). Real-time multi-risk early warning for community banks: An application of ensemble anomaly detection and explainable artificial intelligence. Journal of Advanced Computing Systems.
Lundberg, S., & Lee, S. (2017). A unified approach to interpreting model predictions. Advances in Neural Information Processing Systems.
Mitchell, S., Potash, E., Barocas, S., et al. (2021). Algorithmic fairness: Choices, assumptions, and definitions. Annual Review of Statistics and Its Application.
Najadat, H., Altiti, O., Aqouleh, A., et al. (2020). Credit card fraud detection based on machine and deep learning. 2020 11th International Conference on Information and Communication Systems (ICICS).
Niu, Y., Xiao, H., Liu, D., Chen, N., & Li, J. (2026). Mitigating the safety alignment tax with null-space constrained policy optimization. International Conference on Learning Representations.
Pagano, T., Loureiro, R., Lisboa, F., et al. (2023). Bias and unfairness in machine learning models: A systematic review on datasets, tools, fairness metrics, and identification and mitigation methods. Big Data and Cognitive Computing.
Pamisetty, L., Singireddy, S., Annapareddy, V., et al. (2022). Leveraging AI, machine learning, and big data for enhancing tax compliance, fraud detection, and predictive analytics in government financial management. Migration ….
Pamisetty, V. (2020). Optimizing tax compliance and fraud prevention through intelligent systems: The role of technology in public finance innovation. International Journal on Recent and Innovation Trends in Computing and Communication.
Prakash, P., & Madhavan, A. (2026). Explainable AI and machine learning framework for cyber threat detection and adaptive defense systems. 2026 7th International Conference on Smart Systems and Inventive Technology (ICSSIT).
Qazi, A., Jadoon, A. A. K., & Arshad, S. (2026). APEW-fed: Adaptive privacy-aware ensemble weighting for federated anomaly detection in enterprise resource planning systems. Scientific Reports.
Radhi, M., & Hassan, N. (2026). Automated tax audits and AI-driven compliance: Legal safeguards, transparency, and algorithmic accountability in fiscal administration. Communications in Computer and Information Science.
Scott, W., Richardson, L., Lewis, D., Ward, P., Phillips, S., et al. (2024). Integration of AI and cloud in tax monitoring systems. researchgate.net.
Stanton, C., Katz, G., & Song, D. (2015). Isolation forest for anomaly detection. e3s-center.berkeley.edu.
Stoller, S., Bartocci, E., Seyster, J., Grosu, R., et al. (2011). Runtime verification with state estimation. International Conference on Runtime Verification.
Stradi, F. (2026). Regret minimization in constrained markov decision processes. tesidottorato.depositolegale.it.
Sun, F., Lu, X., Yu, L., & Dong, C. (2026). Adaptive and trustworthy fraud detection for mobile money transactions using PaySim. 2026 6th International Conference on Computer Science and Blockchain (CCSB).
Tan, Y., Zheng, W., Cao, J., & Jiang, B. (2025). Intelligent tax systems: Automating tax audits and improving revenue efficiency. Open J. Account.
Uddin, M. (2026). A regulatory governance framework for AI-driven financial fraud detection in US banking integrating OCC, SR 11-7, CFPB and FinCEN compliance requirements …. Cogent Business & Management.
Wachter, S., Mittelstadt, B., & Russell, C. (2023). Counterfactual explanations without opening the black box. Harvard Journal of Law &, 2023.
Wang, Z. (2024). Adaptive ensemble learning framework with shap-based feature optimization for financial anomaly detection. Journal of Artificial Intelligence Review.
Xiao, X., Chen, H., Zhang, Y., Ren, W., Xu, J., et al. (2025). Anomalous payment behavior detection and risk prediction for SMEs based on LSTM-attention mechanism. Academic Journal of Sociology and Management.
Zhao, R., Shoaib, M., Hoang, V., & Hassan, W. (2025). Rethinking tamper-evident logging: A high-performance, co-designed auditing system. Proceedings of.
Downloads
Published
How to Cite
Issue
Section
Citation Check
License
Copyright (c) 2025 Fahmi Sahlan Fahmi Sahlan, Andi Runis Makkulau, Muhammad Faried Pratama, Muhammad Fery Hidayat
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution-ShareAlike 4.0 International License. In line with the license, authors are allowed to share and adapt the material. In addition, the material must be given appropriate credit, provided with a link to the license, and indicated if changes were made. If authors remix, transform or build upon the material, authors must distribute their contributions under the same license as the original.
