Research on Student Performance Prediction Based on Clustered Graph Neural Networks

The advent of the big data era has brought about profound changes in modern education, making educational data mining an important field within the realm of data analysis. Predicting students' academic performance is one of the crucial topics in this field. This paper proposes a student performance prediction method: Cluster-based Graph Neural Network Prediction (CGNNP). By clustering students' routine and final exam scores and using the clustering results as category labels, the method employs graph neural networks for training and testing the data. Through analyzing students' background information and learning behavior data, this method effectively predicts students' performance. Experimental results demonstrate a significant improvement in prediction accuracy compared to traditional machine learning methods, better reflecting students' learning situations and providing educators with more accurate decision-making support.