Comparison of machine learning algorithms to identify and prevent low back injury

• Autores: Ovalle Paulino, Christian, Huamani Correa, Jorge
• Formato: artículo
• Fecha de Publicación: 2025
• Institución: Universidad Tecnológica del Perú
• Repositorio: UTP-Institucional
• Lenguaje: español
• OAI Identifier: oai:repositorio.utp.edu.pe:20.500.12867/14596
• Enlace del recurso: https://hdl.handle.net/20.500.12867/14596; https://doi.org/10.11591/ijece.v15i1.pp894-907
• Nivel de acceso: acceso abierto
• Materia: Algorithm comparison; Computational medicine; Lumbar injuries; Machine learning; https://purl.org/pe-repo/ocde/ford#2.02.04

With the advancement of technology, remote work and virtual classes have become increasingly common, leading to prolonged periods in front of computers and, consequently, to discomfort and even lower back pain. This study compares machine learning algorithms to identify and prevent low back pain, a common health problem. A predictive model for early diagnosis and prevention of these injuries was developed using datasets from open data repositories. Six machine learning models were used to train the data. Results showed that logistic regression was the most effective model, with performance curves of 70%, 90%, and 99%. Performance metrics indicated 86% accuracy, 85% recall, and 86% F1-score. Accuracy of 70%, recall of 71%, and F1-score of 63% reflect the robust ability of the model to address the problem. In addition, an intuitive interface was implemented using Gradio Software to improve data visualization.