Comparing Automated and Non‐automated Machine Learning for Binary Classification of Running Fatigue Using Wearable Sensors

With the increasing popularity of running, the incidence of running-related injuries has significantly increased. Research indicates that fatigue or overuse is a significant risk factor for running injuries. Therefore, identifying a runner's fatigued state is crucial to prevent injuries. Previous studies have employed manually designed machine learning approaches to achieve global classification for running fatigue using Inertial Measurement Units (IMUs). In this study, our objective is to compare an automated machine learning pipeline with the conventional machine learning approach applied to a benchmark running dataset that achieved up to 75% accuracy. The proposed methodology resulted in an improved performance of up to 6%. These findings demonstrates that automated machine learning has the potential to democratize and enhance accessibility to biomechanics research and applications. This work provides further evidence of the efficacy of automated machine learning and its potential to make a significant impact in the field of sports science.