Spatiotemporal Link Formation Prediction in Social Learning Networks Using Graph Neural Networks

A team from North Carolina State University and the University of North Carolina at Charlotte has developed a graph neural network (GNN) framework that predicts how students will interact in classroom social learning networks (SLNs). SLNs map students as nodes and their interactions as edges, but traditional link prediction methods—designed for general online social networks—fail to capture the complex, non-Euclidean, and dynamically evolving structure of educational settings. The researchers' approach jointly models temporal evolution (how interactions shift over a course) and spatial aggregation (combining data from multiple classrooms) to make more accurate predictions.

Tested on four distinct classrooms, the GNN framework showed statistically significant performance improvements over conventional baselines. Key findings include: prediction accuracy improves as courses progress temporally; aggregating SLNs from multiple classrooms boosts performance, especially in sparse datasets; and jointly leveraging temporal and spatial data significantly outperforms analyzing classrooms in isolation. The work, published at EDM 2026, demonstrates practical value for early-course decision-making and scalable learning analytics, enabling educators to design better group activities and intervene when students are at risk of disengagement.