Analysis of Optimization Effects in Teaching Feedback Mechanisms Driven by Graph Neural Networks in Artificial Intelligence

To enhance the intelligence and structuring of teaching feedback mechanisms, a feedback optimization system based on graph neural networks is constructed. By modeling heterogeneous teaching graphs, multi-relationship representations among students, knowledge points, and exercises are achieved. Combined with feedback score prediction and dynamic path generation algorithms, an adaptively updatable feedback generation framework is formed. Experimental results demonstrate significant improvements over traditional models in feedback accuracy, personalized matching, and path optimization rates, validating the effectiveness of structured graph learning in instructional feedback. This research provides an extensible technical pathway for optimizing feedback in intelligent teaching systems.