Graph Neural Networks for Modeling Knowledge Structures and Optimizing Curricula

Abstract

In the context of the rapid digitalization of educational systems and the exponential growth of available information, there is an increasing need for intelligent methods to structure knowledge and optimize curricula. Traditional approaches to designing educational trajectories often fail to account for the complex interdependencies between disciplines, skills, and competencies, resulting in reduced learning efficiency and insufficient personalization of the educational process. This study proposes the use of Graph Neural Networks (GNNs) as an effective tool for modeling knowledge structures in digital educational systems. The graph - based approach enables the representation of educational content as a network of interconnected elements, where nodes correspond to concepts or learning modules, and edges represent logical, semantic, or pedagogical relationships. Special attention is given to the development of a curriculum optimization model based on the analysis of knowledge graph structures. The proposed method enables the identification of key nodes, optimization of course sequencing, and adaptation of learning trajectories to individual learner needs. The study examines GNN training algorithms, feature aggregation methods, and information propagation mechanisms within graph structures. The experimental results demonstrate that the application of GNNs significantly improves the quality of educational recommendations, enhances curriculum coherence, and reduces redundancy in learning materials. The evaluation shows an increase in learning e fficiency of up to 25% compared to traditional approaches. Thus, Graph Neural Networks represent a promising tool for the intellectualization of educational systems and the development of next-generation adaptive curricula.