A Hybrid Multivariate GRNN–Fuzzy Goal Programming Framework for Fiscal Forecasting and Budget Optimization

Abstract

Economic stability and fiscal sustainability remain fundamental priorities for modern economies, requiring proactive financial management and resilient forecasting frameworks. Traditional linear econometric models often struggle to capture the non-linear dynamics and volatile interactions between government revenues and expenditures. This paper proposes a novel multivariate hybrid framework that integrates Generalized Regression Neural Networks (GRNN) with **Multi-Objective Fuzzy Goal Programming (FGP) for fiscal forecasting and budget optimization under ambiguity. The proposed model captures the dynamic interactions between government revenues and expenditures through a lag-based multivariate structure . In the first stage, the lagged GRNN architecture is deployed to generate reliable, non-linear baseline forecasts for public revenues and expenditures. In the second stage, acknowledging the inherent vagueness and flexible tolerances of real-world budgetary targets, these forecasted trajectories are embedded into an operational layer using Fuzzy Goal Programming. By utilizing linguistic aspiration levels and fuzzy membership functions, the model optimizes budgetary allocations under conflicting and imprecise fiscal constraints, such as flexible deficit reduction and expenditure stabilization. Empirical validation demonstrates that the proposed fuzzy-intelligent hybrid architecture significantly outperforms traditional forecasting benchmarks and provides decision-makers with an adaptive, data-driven optimization tool for resilient fiscal policy and strategic budget planning