Predictive model to prevent injuries in athletes using Machine Learning and sensitivity analysis

Ovalle C.; Sánchez Puche E.; Cerron Valverde W.; Santillán Aching O. · Proceedings of the LACCEI international Multi-conference for Engineering, Education and Technology (2026) · DOI: 10.18687/laccei2026.1.1.412

Resumen

This paper presents the development of a predictive system for the early prevention of muscle injuries in athletes, using Machine Learning and Deep Learning techniques applied to anthropometric and morphological variables. The methodological process was structured based on the CRISP-DM approach, including data cleaning, Min-Max normalization, ordinal encoding, class balancing, and stratified partitioning into training (80%) and test (20%) sets. Six predictive models were evaluated: Random Forest (RF), XGBoost, MultiLayer Perceptron (MLP), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and a Deep Learning model based on artificial neural networks implemented using Keras. Additionally, an Ensemble Voting scheme was developed to integrate individual predictions through majority voting, resulting in a more robust and stable output. Model performance was evaluated using metrics such as Accuracy, Precision, Recall, and F1-score, as well as Precision-Recall curves, allowing for an adequate assessment under class imbalance conditions. The system was subsequently implemented as an interactive web application that enables sports personnel to input athlete data and obtain real-time predictions, facilitating preventive decision-making and contributing to the reduction of muscle injury risk.

Instituciones: Universidad Tecnológica del Perú