Li Yumeng, Koldenhoven Rachel M, Jiwan Nigel C, Zhan Jieyun, Liu Ting
Department of Health and Human Performance, Texas State University, San Marcos, TX, USA.
Department of Kinesiology and Sport Management, Texas Tech University, Lubbock, TX, USA.
Sports Biomech. 2025 Jul 11:1-13. doi: 10.1080/14763141.2025.2528885.
The purpose of the study was to assign rowers to different rowing events based on their demographics and rowing kinematics using machine learning models. A total of 55 elite athletes from the Chinese National Rowing Team participated, each instructed to row on a rowing ergometer for one minute at three stroke rates: 18, 26, and 32 strokes/min. Trunk and upper arm 3D kinematics were collected using an inertia measurement unit system at a sampling rate of 100 Hz. Trunk and upper arm segmental and joint range of motion were generated. Trunk segments and upper arm motion coordination were analysed using the vector coding method. Six supervised machine learning models were trained using the collected demographics and kinematic data to classify rowers' groups (i.e. coxed eight and single/pair event group). The machine learning models successfully classified rowers' groups, with the top-performing models (decision tree, extreme gradient boosting, and random forest) achieving high classification performance (accurate rate = 0.89-0.93). The rowing event assignment automated by machine learning may help coaches make more informed and objective decisions. By minimising subjective biases, this approach enhances the accuracy and fairness of athlete selection processes, thereby potentially optimising team composition and performance outcomes.
该研究的目的是使用机器学习模型，根据划桨运动员的人口统计学特征和划桨运动学特征，将他们分配到不同的划桨项目中。共有55名来自中国国家赛艇队的精英运动员参与其中，每位运动员被要求在赛艇测功仪上以三种划桨频率划桨一分钟，这三种频率分别为：18次/分钟、26次/分钟和32次/分钟。使用惯性测量单元系统以100Hz的采样率收集躯干和上臂的三维运动学数据。生成了躯干和上臂的节段性和关节活动范围。使用矢量编码方法分析了躯干节段和上臂的运动协调性。使用收集到的人口统计学数据和运动学数据训练了六个监督式机器学习模型，以对划桨运动员的组别（即八人有舵手组和单人/双人项目组）进行分类。机器学习模型成功地对划桨运动员的组别进行了分类，表现最佳的模型（决策树、极端梯度提升和随机森林）取得了较高的分类性能（准确率=0.89-0.93）。通过机器学习实现的划桨项目分配自动化可能有助于教练做出更明智、更客观的决策。通过将主观偏差降至最低，这种方法提高了运动员选拔过程的准确性和公平性，从而有可能优化团队组成和成绩。
