Wu Shuying
Liyuan Foreign Language Primary School in Futian District, Shenzhen, 518000, China.
Sci Rep. 2025 Sep 1;15(1):32044. doi: 10.1038/s41598-025-17903-4.
The traditional classroom grade assessment method has some problems, such as strong subjectivity and single dimension, and it is difficult to fully reflect students' real learning state. This study proposes a classroom performance evaluation model based on Graph Convolutional Network (GCN). By constructing an interaction relationship graph among students, this study applies Graph Neural Network (GNN) to educational data analysis to enhance the objectivity and accuracy of evaluation. The study collects multi-source data from teaching management systems, classroom observation records, and online learning platforms. It constructs a graph structure containing students' individual attributes and social relationships, and designs a GCN model architecture and training process suitable for educational scenarios. The experimental results show that the model has achieved significantly better performance than traditional machine learning methods in the four-class classroom performance prediction task. Through ablation experiments and comparative analysis of different graph construction strategies, the important role of social relationship information in student performance prediction is verified. This study not only expands the application path of GNN in the field of educational assessment but also provides new technical ideas for realizing an intelligent and dynamic classroom grade assessment system.
传统的课堂成绩评估方法存在一些问题，如主观性强、维度单一，难以全面反映学生的真实学习状态。本研究提出一种基于图卷积网络（GCN）的课堂表现评估模型。通过构建学生之间的交互关系图，本研究将图神经网络（GNN）应用于教育数据分析，以提高评估的客观性和准确性。该研究从教学管理系统、课堂观察记录和在线学习平台收集多源数据。它构建了一个包含学生个体属性和社会关系的图结构，并设计了适合教育场景的GCN模型架构和训练过程。实验结果表明，在四类课堂表现预测任务中，该模型的性能明显优于传统机器学习方法。通过消融实验和对不同图构建策略的比较分析，验证了社会关系信息在学生成绩预测中的重要作用。本研究不仅拓展了GNN在教育评估领域的应用路径，也为实现智能、动态的课堂成绩评估系统提供了新的技术思路。
