AI-based model improves transparency and efficiency in public finance

By enhancing forecast precision, the system enables governments to preempt budgetary imbalances, reduce waste, and optimize spending efficiency. The authors argue that the combination of machine learning and classical econometrics represents a key evolutionary step toward “smart governance,” where fiscal planning becomes adaptive, evidence-based, and transparent.

A new artificial intelligence–based framework can forecast public spending with higher precision than traditional econometric models. The work, published in Electronics and titled “AI-Driven Models for Forecasting Public Expenditures in the Digital Era,” presents a data-driven approach designed to support financial governance, transparency, and efficiency in public administration.

The study investigates the role of artificial intelligence in improving how governments plan and manage expenditures. Historically, public budget forecasting has relied on econometric models such as the Vector Autoregression (VAR) and Error Correction Model (ECM). While these have proven effective in capturing long-term relationships between revenues and expenditures, they often struggle with nonlinearities and the volatile dynamics of modern fiscal data.

How artificial intelligence improves fiscal forecasting?

To overcome the above limitations, the researchers compared traditional econometric techniques with AI-based models such as the Temporal Convolutional Network (TCN) and Long Short-Term Memory (LSTM). The results revealed that deep learning models could better identify complex temporal patterns and respond more effectively to changes in fiscal behavior caused by external shocks or policy adjustments.

The study used data from 2016 to 2023, drawn from national financial databases, to train and test various forecasting methods. The baseline SARIMA model achieved an average forecasting accuracy of around 88.8%, while the hybrid AI model surpassed 92%, demonstrating measurable gains in predictive reliability. These findings suggest that AI-assisted tools can enhance both short-term and long-term fiscal projections when integrated into public budgeting systems.

Evaluating model performance and economic relationships

The paper also analyzed the structural behavior of fiscal relationships through the ECM approach. The authors found a consistent long-run equilibrium between public revenues and expenditures, estimating that, on average, expenditures tend to adjust by 75% within a one-year period following fiscal imbalances. This means government spending patterns retain a high degree of inertia, yet remain responsive to revenue fluctuations over time.

The machine learning models introduced an additional layer of sophistication. The TCN architecture, in particular, demonstrated robust performance by capturing long-term dependencies in the data. Its convolutional filters allowed for dynamic weighting of sequential data, enhancing its ability to model fiscal seasonality, such as increased spending near year-end and budget reallocation patterns following mid-year policy reviews.

The hybrid framework proposed by the researchers combines the interpretability of econometric models with the adaptability of AI. While SARIMA and ECM retain the advantage of transparent parameter estimation, deep learning components provide the flexibility to process irregular, nonlinear fiscal variations. This integration creates a unified pipeline capable of producing real-time expenditure forecasts that can assist policymakers in scenario planning, resource allocation, and public finance sustainability.

In addition to improved accuracy, the AI-based model achieved lower mean absolute percentage errors (MAPE) and reduced root mean square error (RMSE) values compared with conventional models. These metrics confirm the system’s ability to minimize forecasting deviations, a critical requirement for fiscal authorities that must maintain balanced budgets while adapting to economic uncertainty.

Implications for public governance and digital transformation

According to the researchers, accurate public expenditure forecasting is essential for responsible fiscal governance in the digital era. As governments adopt electronic financial management systems and open-data platforms, the integration of AI into fiscal operations becomes both feasible and necessary. The proposed model supports data-driven decision-making, reduces forecasting bias, and increases the transparency of budgetary processes.

The study also aligns with broader European digital transformation goals, promoting automation and predictive analytics within public sector management. By enhancing forecast precision, the system enables governments to preempt budgetary imbalances, reduce waste, and optimize spending efficiency. The authors argue that the combination of machine learning and classical econometrics represents a key evolutionary step toward “smart governance,” where fiscal planning becomes adaptive, evidence-based, and transparent.

Moreover, the research highlights the role of AI in supporting risk assessment and fiscal resilience. As economic shocks, such as inflation surges or geopolitical disruptions, become more frequent, predictive models that incorporate adaptive algorithms can provide early warnings, allowing fiscal institutions to react before crises deepen. This capability could be particularly valuable for ministries of finance, central banks, and audit bodies managing multi-level budgets across different government agencies.

Another significant contribution of the paper lies in its methodological reproducibility. The authors provide a clear implementation roadmap for integrating AI forecasting systems within existing public finance infrastructures. The model can operate alongside current financial management information systems (FMIS) without disrupting established workflows, facilitating gradual adoption across public institutions.

The road ahead: Toward AI-driven fiscal intelligence

While the study demonstrates clear advantages of AI-based forecasting, the authors acknowledge challenges related to data quality, interpretability, and institutional readiness. Public datasets often contain inconsistencies, delays, and missing values that can reduce model robustness. To address these limitations, the research recommends continuous data curation and hybrid modeling, combining interpretable econometric equations with deep learning layers trained on cleaned, verified data.

The team also calls for interdisciplinary collaboration among data scientists, economists, and policymakers to ensure responsible implementation. Transparency in AI decision-making remains a priority, especially in fiscal contexts where algorithmic forecasts directly influence public resource distribution.

Looking ahead, the study proposes extending the framework with federated learning and explainable AI (XAI) approaches to enhance privacy and interpretability. These methods would enable multiple government agencies to collaboratively train forecasting models without sharing sensitive financial data, fostering national and regional cooperation in fiscal management.