A novel machine learning framework for predicting and preventing neck pain in mobile device users through ergonomic optimization

Prolonged use of mobile devices and poor posture have significantly increased the risk of neck discomfort among users. To address this concern, ergonomic optimization combined with advanced machine learning can help identify and mitigate musculoskeletal risks. However, accurately predicting individual ergonomic risk factors remains challenging due to varied device interactions, usage behaviors, and anatomical differences. This study proposes a novel machine learning framework for predicting and preventing neck pain in mobile device users by quantitatively forecasting a neck strain risk score based on sensor-derived ergonomic features. The approach utilizes the NeckCheck dataset, applies notch filtering for data pre-processing, and extracts features using the Histogram of Oriented Gradients for the neck region. Prediction is performed using an Enhanced Support Vector Regression (ESVR) model, with parameters tuned by the Shuffled Shepherd Optimization Algorithm (SSOA) to minimize error and maximize accuracy. Simulation results demonstrate that the proposed framework achieves superior prediction accuracy and lower root mean square error (RMSE) compared to conventional models, supporting its effectiveness in reducing ergonomic risk and preventing neck pain. This work presents a methodological framework for estimating neck-related ergonomic risk using motion-based features and optimization-enhanced regression modelling. The study focuses exclusively on algorithmic development and computational validation using the NeckCheck dataset, without involving direct clinical trials or physician-verified assessments. Accordingly, the findings should be interpreted as model-level evidence rather than clinical diagnostic validation.