Classification and Performance Modeling of Students in Long Jump Using Machine Learning

Abstract

This study evaluates the effectiveness of machine learning models in classifying students’ long jump performance into distinct categories using physiological and performance metrics. Five classification algorithms—K-Nearest Neighbors (KNN), Random Forest, XGBoost, Gradient Boosting, and Support Vector Machine (SVM)—were assessed for their predictive accuracy and overall performance. The KNN model achieved the highest accuracy of 91.5%, followed closely by the Random Forest classifier at 90.1%, demonstrating strong performance in precision (0.916 and 0.901), recall (0.915 and 0.901), and F1 score (0.915 and 0.901), respectively. The XGBoost and Gradient Boosting models provided competitive results with accuracies of 89.6% and 87.3%, showing balanced performance across all metrics. Conversely, the SVM model exhibited the lowest accuracy at 68.4%, with relatively lower precision (0.689), recall (0.684), and F1 score (0.685).The confusion matrices revealed that KNN and Random Forest models had the fewest misclassifications, particularly excelling in distinguishing between the four performance categories. The study highlights that while KNN offers the highest accuracy, Random Forest provides greater stability and generalization, making both models suitable for practical applications in sports analytics. These findings underscore the potential of machine learning techniques in supporting physical education teachers, coaches and trainers by enabling data-driven decisions to optimize training programs and enhance long jump performance among youth athletes.