Graph structure modeling for optimizing the relationship between training load and physical health in volleyball players using Graph of Thought

To address the challenge of aligning training-load regulation with individual physiological feedback in volleyball players, this study proposes a personalized optimization framework based on Graph of Thoughts (GOT). An individualized knowledge graph is first constructed as a domain prior, and real-time monitoring data are then used to drive a four-stage GOT reasoning process comprising perception, analysis, prediction, and optimization. A residual-triggered update mechanism continuously adjusts graph nodes and edge weights to improve model adaptability. In addition, Analytic Hierarchy Process (AHP)-based weighting is introduced to generate personalized training plans that balance performance enhancement with health-risk prevention, while a near-real-time closed-loop feedback system supports dynamic intervention. Experimental results show that the proposed method achieved mean consistency scores of 0.89 ± 0.04 for heart rate variability (HRV) prediction and 0.86 ± 0.06 for creatine kinase (CK) prediction. The framework also reduced injury risk and improved overall physical health, with the comprehensive health index reaching 0.71. These findings suggest that the proposed approach shifts training-load modeling from static group-level analysis to dynamic individualized reasoning, providing an interpretable and iterative solution for intelligent training management.