Application of Inertial Measurement Units and Machine Learning for Fatigue Assessment in Badminton Athletes

Objective : The aim of this study was to develop and evaluate binary classification models to detect physical fatigue in badminton athletes using inertial measurement unit (IMU) data and machine learning algorithms. Methods : Thirty-two collegiate badminton athletes participated in this study. Movement data of these participants were collected using multi-sensor IMUs placed on key body regions and a single forearm IMU sensor before and after fatigue induction. Feature selection was performed using Lasso regression to identify the most relevant kinematic features. Six machine learning models—support vector machine (SVM), random forest (RF), logistic regression (LR), k-nearest neighbors (KNN), decision tree (DT), and naïve Bayes (NB)—were applied to construct binary classification models to detect fatigue. Models’ performances were evaluated based on area under the curve (AUC), accuracy, sensitivity, precision, and F1 score. Results : Overall performance of the SVM model was the best across both multi-sensor and single-sensor configurations, with AUC values of 0.89 and 0.95. The single-sensor forearm IMU model achieved high predictive accuracy, underscoring the feasibility of using simplified setups for fatigue monitoring. Feature selection by Lasso regression revealed key fatigue indicators, including forearm kinematic features, contributing significantly to model accuracy. Conclusion : IMU data combined with machine learning models can reliably be used to assess physical fatigue in badminton athletes. High performances of multi-sensor and single-sensor configurations suggest flexibility in model application, supporting real-time, field-based monitoring for fatigue management, performance optimization, and injury prevention in sports. Future research should validate these models in broader athletic populations and explore additional data sources to enhance their predictive abilities.