Research on Knowledge Defect Detection and Performance Prediction Algorithm Based on Deep Structure Learning and Time Feature Extraction Method

This study explores the knowledge deficiency and performance prediction algorithm based on deep structure learning and temporal feature extraction, and proposes an innovative model. The model innovatively integrates deep learning technology, especially through the attention-driven RGG strategy, which integrates the attention mechanism at the feature level and focuses on the key learning influencing factors. It embeds the attention mechanism into R-GCN to deepen the understanding and analysis of the student relationship network. At the same time, it uses GRU network to capture student behavior patterns evolving over time, and uses the memory function of GRU to extract key behavior features. The joint application of R-GCN and GRU gives full play to the advantages of both, greatly improving the prediction efficiency and classification accuracy. In order to verify the effect of the model, the OULAD dataset was selected to compare with a variety of traditional and deep learning algorithms, and ablation experiments were carried out to analyze the role of each module. Experiments show that the model performs well in binary classification prediction, accurately identifies students with learning difficulties, provides a basis for personalized intervention, and effectively promotes learning monitoring and support.