Construction and precise profiling of a holistic student evaluation system based on multimodal data integration

Graph Neural Networks (GNNs) offer an effective solution for integrating multi-modal educational data, including academic records, behavioral engagement logs, and social interaction patterns. This study introduces a GNN-based holistic student evaluation framework that combines advanced data preprocessing techniques such as normalization, dimensionality reduction, and imputation to handle heterogeneous and incomplete datasets. By modeling students as graph nodes and their relationships as edges, the framework captures both individual attributes and relational dependencies. Experiments conducted on a dataset with 50,000 nodes and 300,000 edges demonstrated the framework’s scalability and effectiveness. The model achieved a classification accuracy of 92.4% and an F1-score of 92.4%, significantly outperforming traditional machine learning models, while maintaining a real-time inference speed of 18ms per sample. With the rapid development of smart cities and smart campuses, the proposed framework provides a crucial technical foundation for intelligent educational governance, enabling real-time, personalized student services and supporting the construction of a data-driven smart learning ecosystem. These results highlight the capability of the framework to provide precise student profiling and actionable insights into academic, behavioral, and social dimensions, addressing the limitations of traditional evaluation methods and supporting data-driven educational decision-making.