Supervised learning techniques and their ability to classify a change of direction task strategy using kinematic and...

Area of Science:

• Biomechanics
• Sports Science
• Machine Learning

Background:

• Athletic groin pain affects numerous male athletes.
• Understanding movement patterns is crucial for injury prevention and rehabilitation.
• Previous research categorized subjects into distinct movement strategies.

Purpose of the Study:

• To evaluate supervised learning techniques for classifying change of direction (COD) tasks.
• To assess the impact of fuzzy executions and specific biomechanical features on classification accuracy.
• To identify the most influential features for movement pattern classification.

Main Methods:

• Utilized kinematic and kinetic data from 323 male subjects with chronic athletic groin pain.
• Applied various supervised learning algorithms including decision trees, support vector machines, and neural networks.
• Analyzed features such as hip/pelvis angles, thorax sway, and ankle moments.

Main Results:

• The correlation approach (82%) and support vector machine (80%) showed the highest classification accuracy.
• Fuzzy observations ranged from 16% to 25% of the data.
• Key features included hip flexion, ankle rotation, thorax sway, and hip/pelvis abduction angles.

Conclusions:

• Supervised learning models can predict movement patterns with high accuracy (~84%).
• The findings support the existence of multiple movement patterns for COD tasks.
• Specific biomechanical features significantly influence the classification of movement strategies.