A combined approach for predicting elbow joint moment and muscle force during dumbbell kickback in females using simulation-driven machine learning and metaheuristic optimization.

This study evaluates metaheuristic hyperparameter optimization for estimating elbow joint moment (EJM) and triceps brachii muscle force (TBMF) during dumbbell kickback in female participants. Motion data from fourteen women were used in a MATLAB Simscape Multibody link-segment model to generate biomechanical outputs. SVR and DTR models were trained and optimized using AVOA, ChOA, GTO, and HHO. Optimization markedly improved prediction accuracy, with GTO-optimized SVR achieving the best performance (R² = 0.9891 for EJM and 0.9940 for TBMF). The framework supports efficient estimation of joint moments and muscle forces from limited kinematic inputs.