Dynamic Model for Budget Allocation in via Multi-Criteria Optimization

Abstract

This research introduces a dynamic multi-criteria optimization framework for fair budget distribution across four districts in Kazakhstan’s Almaty region. Its main objective is to promote transparency, equity, and efficiency in allocating a constrained regional budget of 42,656,543 thousand tenge across seven activity areas (AA): education, healthcare, transport, infrastructure, digitalization, culture, and ecology. The framework incorporates four weighted criteria: citizen satisfaction (0.2 weight), strategic development priorities (0.2), basic needs fulfillment (0.3), and urbanization level (0.3). Two optimization techniques were employed: Sequential Quadratic Programming (SQP) in MATLAB, converging in 100 iterations with an objective function value of 18,519,864.85 thousand tenge, and Genetic Algorithm (GA) in Python, achieving a slightly higher value of 18,520,000.00 thousand tenge after 500 generations. The minimal difference of 135.15 thousand tenge (0.0007% of the budget) underscores the reliability of both methods. All seven sectors received funding, with healthcare (22.05%) and transport (21.11%) allocated the largest portions, and education (7.03%) the smallest. Fairness is evidenced by a standard deviation of sectoral shares at 5.69%, a coefficient of variation of 0.398, and a Gini coefficient of 0.223. Participatory budgeting was simulated using synthetic citizen voting data derived from demographic factors. Visualizations depict the optimization process’s convergence and budget distribution across feasible solutions. A proposal for pilot testing within Kazakhstan’s e-government system (Egov) has been submitted to the Ministry of Digital Development. Future enhancements will include explainable AI, stakeholder-driven weight adjustments, and real demographic and budgetary data to foster transparency and public confidence. This framework provides a scalable, data-driven approach to participatory budgeting, harmonizing strategic objectives, socio-economic demands, and citizen preferences. SQP and GA methods achieved near-optimal solutions with objective function values of 18,519,864.85 and 18,520,000.00 thousand tenge, respectively. The 135.15 thousand tenge difference (0.0007% of the budget) is negligible, confirming their robustness.