Uncovering Social Dynamics in Education Through Graph Neural Networks and Continual Learning Methods

The informal structure of classrooms forces researchers to contemplate the social aspects that shape student behavior. Such models have been useful despite their omission of connections between students. Aimed at resolving this gap, graph neural networks and continual learning are introduced to represent and interact with academic networks and find correlations affecting academic performance. The proposed work, which includes monitoring student behavior, attendance, participation, and online metrics, will be utilized with a, and online activities, with a Graph Convolutional Network (GCN) model for social network mapping. The scope of this research evolves as technology and educational policies change, thanks to continual learning. Comparing the performance of the proposed model with models such as DGNN, LSTM-NN, ASSAN have shown better accuracy. The proposed model is better than existing ones with an accuracy of 86.50%, 81.30% F1-score, and a precision of 84.35% at 100% data saturation. This becomes useful in complementing research by helping educators to recognize the social factors and plan in a way that positively influences the learning outcomes of the students.