Abstract
This study proposes a hybrid machine learning framework that integrates one-dimensional convolutional neural networks (1D-CNN) with multi-head attention and Light Gradient Boosting Machines (LightGBM) to model the relationship between physical fitness and body mass index (BMI), thereby generating personalized exercise prescriptions. The dataset consists of 6,698 male students aged 18–20 years, including BMI measurements alongside four standardized fitness indicators: 3,000-meter run (aerobic capacity), pull-up test(muscular strength), sit-up test (muscular endurance), and 30 × 2 shuttle run (anaerobic capacity). The 1D-CNN + Attention module effectively captures both local and global temporal patterns, while LightGBM significantly enhances classification accuracy through gradient-boosted decision trees. The proposed hybrid architecture achieved state-of-the-art performance in BMI classification, with an accuracy of 94.5% (Cohen’s κ = 0.91) and an F1 score of 0.93, outperforming traditional classifiers by 12.3% to 19.1%. Model interpretability is ensured through SHapley Additive exPlanations (SHAP), which supports dynamic prescription adjustments aimed at improving muscular strength, cardiorespiratory endurance, speed, agility, and flexibility. A 12-week randomized trial demonstrated the clinical efficacy of this framework, yielding a 23.5% reduction in overweight and obesity prevalence, a 15.2% increase in pull-up test performance, and a 9.8% improvement in 30 × 2 shuttle run results. With an inference time of less than 0.8 milliseconds per sample and robust clinical outcomes, this framework provides a scalable real-time solution for data-driven health optimization. It’s well-suited for both clinical and mobile healthcare applications, addressing the growing demand for personalized exercise interventions among young adults.
Data availability
The datasets generated and/or analysed during the current study are not publicly available due to institutional data governance policies and the need to protect participant privacy inthis homogeneous student cohort, but are available from the corresponding author on reasonable request.
References
World Health Organization. Obesity and overweight: Key facts. (2021). https://www.who.int/news-room/factsheets/detail/obesity-and-overweight.
Smith, J. D., Fu, E. & Kobayashi, M. A. Prevention and management of childhood obesity and its psychological and health comorbidities. J. Am. Med. Assoc. 323 (16), 1612–1614. https://doi.org/10.1001/jama.2020.1410 (2020).
Bhaskaran, K., Dos-Santos-Silva, I., Leon, D. A., Douglas, I. J. & Smeeth, L. Association of BMI with overall and cause-specific mortality: A population-based cohort study of 3.6 million adults in the UK. Lancet Diabetes Endocrinol. 6 (12), 944–953. https://doi.org/10.1016/S2213-8587(18)30288-2 (2018).
Ortega, F. B., Ruiz, J. R., Castillo, M. J. & Sjöström, M. Physical fitness in childhood and adolescence: A powerful marker of health. Int. J. Obes. 32 (1), 1–11. https://doi.org/10.1038/sj.ijo.0803774 (2008).
Warburton, D. E. & Bredin, S. S. Health benefits of physical activity: A systematic review of cu rrent systematic reviews. Curr. Opin. Cardiol. 34 (5), 541–556. https://doi.org/10.1097/HCO.0000000000000640 (2019).
Topol, E. J. High-performance medicine: The convergence of human and artificial intelligence. Nat. ure Med. 25 (1), 44–56. https://doi.org/10.1038/s41591-018-0300-7 (2019).
Liang, J., Matheson, B. E., Kaye, W. H. & Boutelle, K. N. Neurocognitive correlates of obesity and obesity-related behaviors in children and adolescents. Int. J. Obes. 39 (4), 494–506. https://doi.org/10.1038/ijo.2014.170 (2015).
Althnian, A., Aloboud, N., Aldhafian, O. & Alharbi, S. LSTM-based model for BMI trajectory pr ediction in Saudi Arabia. IEEE Access. 9, 96298–96308. https://doi.org/10.1109/ACCESS.2021.3094207 (2021).
Chen, T. & Guestrin, C. XGBoost: A scalable tree boosting system. In Proceedings of the 22nd AC M SIGKDD International Conference on Knowledge Discovery and Data Mining, 785–794. https://doi.org/10.1145/2939672.2939785 (2016).
Ronao, C. A. & Cho, S. B. Human activity recognition with smartphone sensors using deep learn ing neural networks. Expert Syst. Appl. 59, 235–244. https://doi.org/10.1016/j.eswa.2016.04.0 (2016).
Nes, B. M. et al. Estimating VO₂peak from a nonexercise prediction model: The HUNT Study, Norway. Med. Sci. Sports Exerc. 43 (12), 2354–2360. https://doi.org/10.1249/MSS.0b013e318223ac56 (2011).
Martínez-Rodríguez, A., Sánchez-Sánchez, J., Martínez-Olcina, M. & Vicente-Martínez, M. Reinfor cement learning for resistance training recommendations. Appl. Sci. 11 (12), 5393. https://doi.org/10.3390/app11125393 (2021).
Ahmadi, M. N., Pfeiffer, K. A. & Trost, S. G. Bayesian optimization of HIIT interventions using heart rate variability. J. Sci. Med. Sport. 23 (12), 1156–1161. https://doi.org/10.1016/j.jsams.2020.05.019 6/j.jsams.2020.05.019 (2020).
Jurca, R. et al. Associati on of muscular strength with incidence of metabolic syndrome in men. Med. Sci. Sports Exerc. 37 (11), 1849–1855. https://doi.org/10.1249/01.mss.0000175865.17614.8a (2005).
Patel, M. S., Volpp, K. G. & Asch, D. A. Nudge units to improve the delivery of health care. N. Engl. J. Med. 381 (22), 2103–2105. https://doi.org/10.1056/NEJMp1906489 (2019).
Funding
Supported by Natural Science Foundation of Hunan Province (Grant No.: 2024JJ8032) and the Hunan Social Science Achievement Evaluation Committee (Grant No.: XSP25YBC513).
Author information
Authors and Affiliations
Contributions
Conceptualization and validation: Ye Yang; Methodology: Ming Mo; Formal analysis: Guixiang Wu; Writing–original draft: Buxi Li; Writing–review & editing: Peng Kang; Supervision: Jun Wang, Xuyin Xu; Project administration: Wanhong Luo, Tianshuo Jiao.All authors read and approved the final version of the manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Ethics statement
This study involved two phases. The model development phase utilized retrospective anonymized physical fitness data from the model-development cohort of 6,698 male students (aged 18–20) at Changsha Aeronautical Vocational and Technical College, Hunan province, China, collected from 2020 to 2024. The data, acquired through the institution’s secure platform on April 1, 2024, included anthropometric measurements (height, weight, BMI) and physical fitness test data (3,000-meter run, 30 × 2 shuttle run, pull-up test, sit-up test). A two-stage anonymization protocol was implemented: direct identifiers (name and student ID) were removed at the data collection stage, and numerical perturbation (height ± 0.5 cm; weight ± 1 kg) was applied to prevent re-identification. The intervention validation phase was a prospective randomized controlled trial with the intervention cohort of 1,160 participants approved by the Research Office of Changsha Aeronautical Vocational and Technical College. Informed consent was obtained from all participants in the trial. The study was conducted in accordance with the Declaration of Helsinki. Informed consent was obtained from all participants, and ethical approval was granted by the Human Ethics Committee of Chizhou University (Approval No.: 202301015).
Reporting guidelines
This study adheres to the CONSORT (Consolidated Standards of Reporting Trials) checklist (for non-randomized controlled trials).
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Below is the link to the electronic supplementary material.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Mo, M., Li, B., Yang, Y. et al. A machine learning framework for personalized exercise prescription based on BMI and physical fitness assessment. Sci Rep (2026). https://doi.org/10.1038/s41598-026-42405-2
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-026-42405-2
