Classification of musculoskeletal pain using machine learning

Musculoskeletal pain is a significant health concern affecting individuals across various demographics and professions, often leading to reduced productivity and impaired quality of life. This study proposes a framework leveraging Particle Swarm Optimization (PSO) to evaluate and assess musculoskeletal pain risk based on a comprehensive dataset encompassing demographic, professional, physical, and lifestyle characteristics. The dataset includes detailed information on individuals' pain experiences across multiple body regions, providing a robust foundation for identifying correlations and risk factors. By integrating PSO with neural networks, this framework aims to enhance the detection of pain risk patterns, offering insights into the interplay between various factors and musculoskeletal health. The proposed framework involves data preprocessing, definition of neural network architecture, implementation of PSO, and performance evaluation. The dataset, containing 350 entries, was preprocessed to handle missing values, balance class distributions using SMOTE, and normalize features. A fully connected feedforward neural network with a single hidden layer was employed, with PSO optimizing the network's weights and biases. Performance was evaluated using metrics including accuracy, precision, recall, F1-score, and AUC-ROC. The results demonstrate that the PSO-optimized neural network effectively identifies musculoskeletal pain risk, achieving strong performance across all evaluation metrics (accuracy 95.8-100%). Key determinants such as age, BMI, exercise frequency, and occupational factors were identified, providing valuable insights for targeted interventions. The framework's performance compares favorably with conventional approaches, highlighting the potential of optimization techniques in musculoskeletal pain assessment and the development of preventive strategies.