Data-Driven risk assessment framework for federal grant accountability in U.S. state-administered social service and community development programs
Finance & Accounting Research Journal · 24.03.2026 · 8 (3) · 106-123
DOI: 10.51594/farj.v8i3.2234 ↗Kurz gefasst
Fachartikel in der Zeitschrift Finance & Accounting Research Journal, online erschienen am 24.03.2026. Verfasst von Eric Acheampong und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).
Abstract (Originalsprache)
Data-driven risk assessment frameworks for federal grant oversight are increasingly vital to mitigating financial misappropriation and ensuring the equitable distribution of resources across state-administered social programs. The study explores the efficacy of ensemble machine learning techniques in identifying anomalous patterns within grant expenditures, contrasting traditional econometric benchmarks with advanced algorithmic architectures to improve the precision of federal oversight. The study used a systematic literature review to categorize key financial and behavioral indicators, subsequently training models like XGBoost and Random Forest on integrated expenditure datasets to enhance the detection of high-risk transactions. These models demonstrate high predictive sensitivity by leveraging behavioral indicators to flag potential non-compliance before the disbursement of state-administered funds. The findings of the study revealed that gradient boosting algorithms significantly outperform standard logistic regression in identifying irregular financial patterns, achieving higher recall rates that are essential for preemptive intervention in complex grant allocation cycles. Empirically, the results showed that the integration of ensemble architectures effectively mitigated overfitting and increased the precision of risk flagging by 15% compared to baseline models. The results also revealed that incorporating SMOTE balancing techniques further enhances the model's ability to detect fraudulent activity within imbalanced datasets, effectively addressing the minority class challenges inherent in federal expenditure reporting. Also, the study emphasizes the importance of interdisciplinary regulatory collaboration to ensure that algorithmic outputs are effectively integrated into existing federal oversight workflows. The study, therefore, concludes that the implementation of machine learning-based risk stratification serves as a scalable solution for closing oversight gaps, providing federal agencies with actionable intelligence to prioritize audits and preserve the integrity of community development initiatives. Keywords: Federal Grant Accountability, Xgboost, Machine Learning, Risk Assessment, Expenditure Anomaly Detection, Predictive Oversight.
Zitieren
APA
Acheampong, E., & Adukpo, T. K. (2026). Data-Driven risk assessment framework for federal grant accountability in U.S. state-administered social service and community development programs. Finance & Accounting Research Journal, 8(3), 106-123. https://doi.org/10.51594/farj.v8i3.2234
Vancouver
Acheampong E, Adukpo TK. Data-Driven risk assessment framework for federal grant accountability in U.S. state-administered social service and community development programs. Finance & Accounting Research Journal. 2026;8(3):106-123. doi:10.51594/farj.v8i3.2234
BibTeX
@article{Acheampong2026,
author = {Acheampong, Eric and Adukpo, Tobias Kwame},
title = {Data-Driven risk assessment framework for federal grant accountability in U.S. state-administered social service and community development programs},
journal = {Finance & Accounting Research Journal},
year = {2026},
volume = {8},
number = {3},
pages = {106--123},
doi = {10.51594/farj.v8i3.2234}
}
EUVIMED