<p>This study proposes a Feedback Evaluation System using Knowledge Transfer Learning (FES-KTL) to enhance the effectiveness of physical education teaching by systematically analyzing student feedback. The primary objective is to improve performance-centric teaching strategies through automated evaluation and classification of feedback into constructive or modifiable categories. The proposed method employs a knowledge transfer learning framework to align historical performance data with current feedback, iteratively optimizing feature relevance and classification accuracy. The model incorporates external performance metrics and training mode data to refine the classification process and adaptively learn from evolving feedback patterns. Experimental results demonstrate that FES-KTL significantly improves feedback classification accuracy compared to baseline models, with consistent performance across multiple teaching modes and student groups. This work contributes to the field by introducing an AI-driven mechanism for real-time teaching assessment, promoting data-informed pedagogical adjustments, and offering a scalable approach applicable to broader educational contexts.</p>