Machine learning methods to classify engineering students’ ability to perform recreational archery activities using biomechanical data

Utilizing machine learning approaches to predict students’ performance is a valuable tool for anticipating both low and high levels of achievement across different educational levels. In the case of engineering programmes, the students must exhibit proficient hands-on abilities that match their knowledge, to excel in their performance. When assessing the psychomotor abilities of engineering students during hands-on, practical, or laboratory tasks, the available assessment tools are limited. Utilizing both biomechanical data and observations of students’ performance provides an alternate method for evaluating psychomotor skills. This study introduced a recreational archery game as a proof of concept to examine the impact of engineering students’ biomechanical data-the upper-limb physical attributes, posture, and angle-on their performance as measured by the points they score in the archery activity. We recruited 104 engineering students without prior experience performing recreational archery and conducted the activity in laboratory settings. We developed a classification model combining fourteen predictors to categorize the students into two groups: those able to perform and those who are not. The model was run into three different machine learning classification algorithms including the decision tree, K-nearest neighbour, and support vector machine (SVM). Among all preset models, SVM shows the best accuracy, and then further optimization of the SVM model was conducted to achieve higher prediction model accuracy. The optimization of the SVM model increased the testing accuracy to 80 % with F1 scores of 0.75 and 0.83 in predicting the ‘not able to perform’ and ‘able to perform’ classes respectively. Thus, this study shows that a combination of biomechanical data and machine learning is a potential tool for assessing engineering students’ performance involving their psychomotor abilities.