Multi-Agent System for planning the educational contingent using neural networks

Abstract

This paper is dedicated to the development and validation of an intelligent architecture for educational cohort planning, based on the integration of multi-agent systems with artificial neural networks. The research is motivated by the need to efficiently manage educational data flows characterized by high volume, temporal dynamics, and uncertainty, in the context of demographic and socio-economic changes. To this end, a formal model is proposed that describes agents’ decision-making dynamics, inter-agent coordination mechanisms, and the neural learning process used to predict key educational indicators. To validate the proposed solution, an experimental dataset covering the period 2020–2024 was used, reflecting the educational trajectory from high school graduation to the completion of undergraduate studies. The experimental results highlight stable convergence of the learning process, a reduction in prediction error, and the model’s ability to approximate nonlinear relationships between demographic and socio-economic factors and educational indicators. The multi-agent architecture enables efficient distribution of computational tasks, scalability, and adaptability to changes in the educational environment. The proposed solution provides robust decision support for educational management and may serve as an essential formal basis for the development of advanced intelligent systems for institutional planning.