A graph neural network-based decision support model for educational management

Educational management increasingly incorporates intelligent analytics to interpret complex academic and behavioral information for timely decision-making. Existing decision support approaches often focus on limited performance indicators and are not designed to address broader educational management needs such as engagement tracking, progression analysis, and resource alignment. This research aims to develop a graph neural network-based decision support model for educational management using a novel Dynamic Ant Colony-mutated Spatial–Temporal Graph Neural Network (DAC-ST-GNN). The dataset comprises 1000 student records containing academic scores, attendance history, digital activity traces, participation logs, and course-related information collected from institutional repositories and learning platforms. Preprocessing employs Min–Max normalization to ensure consistent and noise-free data. The proposed framework encodes learners and educational entities as graph nodes, with temporal and contextual dependencies forming spatiotemporal edges. DAC-ST-GNN integrates a DAC optimization strategy to iteratively refine graph weights, with ST modeling capturing both relational and time-based dependencies. The GNN framework then learns complex interactions among nodes and edges for enhanced educational data representation and prediction. Implemented in Python, the model delivers actionable insights for administrators and educators, including performance forecasts, engagement alerts, and recommendations. Experimental evaluation demonstrates 98.6% accuracy, 98.5% recall, 98.2% precision, and 98.1% F1-score, along with the lowest RMSE of 37.82 and MAPE of 26.03, outperforming baseline GNN approaches. These results indicate that DAC-ST-GNN provides a highly reliable, adaptive, and effective decision support mechanism for informed educational planning and academic enhancement.