Heterogeneous graph data mining and evaluation system optimization for educational quality based on graph convolutional networks
Traditional educational quality evaluation systems rely on single-dimensional static indicators and fail to capture complex interdependencies among educational entities, leading to one-sided and lagged assessments. This paper proposes an optimized educational quality evaluation system based on Graph Convolutional Networks (GCNs) to address the graph-structured and multi-source heterogeneous nature of educational data. We first construct a heterogeneous graph model integrating students, teachers, courses, and school facilities. A hybrid data-driven and knowledge-driven GCN framework with multi-head attention is then designed to enhance feature representation while ensuring data privacy via federated learning. Experiments on 8-year multi-dimensional data from a middle school show that the proposed method achieves 89.7%, 92.5%, and 91.1% accuracy on student performance prediction, teacher evaluation, and school quality assessment, respectively, outperforming traditional methods by 8.3%-14.6% and reducing MAE (Mean Absolute Error) by over 17%. This approach provides scientific decision support for educational administrators. The final HEQ-GCN is evaluated with a chronological split comprising 2015–2019 for training, 2020 for validation, and 2021–2022 for testing, and all three reported accuracies are obtained from the same unified model.
Authors
- Yu Wang
- Yanling Wang
Institutions
- Qiqihar University (CN)
Publication Details
- Journal
- Discover Artificial Intelligence
- Published
- 2026-11-21
- DOI
- https://doi.org/10.1007/s44163-026-02397-4
- Primary Topic
- Advanced Graph Neural Networks
- Type
- article
- Field-Weighted Citation Impact
- 0.00