DETERMINATION OF BODY MASS INDEX USING MACHINE LEARNING REGRESSION METHODS FROM BODY COMPOSITION PARAMETERS
SERTEL MEYVACI S 1,
SECGIN Y 2,
ULGER T G 3,
TASLAMACIOGLU DUMAN T 4,
CELIK B 1,
DEMIROGLU S 1 * More Detail
1 Department of Anatomy, Faculty of Medicine, Bolu Abant Izzet Baysal University, Bolu, Türkiye
2 Department of Anatomy, Faculty of Medicine, Karabük University, Karabük, Türkiye
3 Department of Nutrition and Dietetics, Faculty of Health Sciences, Bolu Abant Izzet Baysal University, Bolu, Türkiye
4 Department of Internal Medicine, Faculty of Medicine, Bolu Abant Izzet Baysal University, Bolu, Türkiye
* Corresponding Author
J CLIN MED KAZ, In press.
OPEN ACCESS
85 Views
0 Downloads
ABSTRACT
Introducton: Obesity is one of the most significant global health problems increasing today. Body mass index (BMI) is a fundamental parameter widely used to assess obesity. However, evaluating different measurements of body composition may contribute to determining BMI more accurately and comprehensively. In this context, machine learning (ML) provides a powerful alternative to classical methods by predicting BMI from body composition. Based on this hypothesis, the aim of this study is to predict BMI using ML regression models from body composition parameters.
Methods: The study included 411 individuals aged 18–65 years. The individuals’ body weight and body composition parameters [fat mass index, body fluid percentage, fat mass (kg), fat-free mass (kg), body fat percentage and fat-free mass percentage] were measured using the Tanita MC 580 body composition analyzer. Data were analyzed using 16 different ML regression models with the Python language and PyCaret library. Model success was expressed by the R2 Comparison value.
Results: The highest success was achieved with Linear Regression, Bayesian Ridge, and Ridge Regression models (R²=0.9937). Huber Regressor (0.9928) and Least Angle Regression (0.9911) followed. Among the models with low success was K Neighbors Regressor (0.8316). In the parameter analysis, Fat-Mass Index was the strongest predictor in BMI estimation.
Conclusion: This study shows that ML-based regression models can predict BMI with high accuracy using body composition parameters, with the fat mass index providing the most significant contribution. These results emphasize the potential of digital phenotyping as a valuable approach in obesity research and clinical evaluation. By framing BMI not only as a traditional measurement but as a more comprehensive health marker supported by digital tools, the findings point toward the development of personalized, rapid, and reliable digital health solutions for obesity screening and monitoring.
Keywords: Body mass index, obesity, machine learning regression models, bioelectrical impedance analysis, digital phenotyping
CITATION
S SM, Y S, G UT, T TD, B C, S D. DETERMINATION OF BODY MASS INDEX USING MACHINE LEARNING REGRESSION METHODS FROM BODY COMPOSITION PARAMETERS. J Clin Med Kaz. 2026.