Machine learning approach to predict body weight in adults

Area of Science:

• Biomedical Informatics
• Machine Learning in Healthcare
• Predictive Modeling for Health Outcomes

Background:

• Obesity is a major risk factor for non-communicable diseases like type 2 diabetes, hypertension, and cardiovascular disease.
• Effective weight control is crucial for preventing these chronic conditions.
• A rapid method to predict future weight changes is needed for clinical weight management.

Purpose of the Study:

• To evaluate a machine learning model's capability in predicting body weight changes over a three-year period.
• To utilize big data and advanced algorithms for accurate weight prediction.
• To develop a tool for proactive and personalized weight management.

Main Methods:

• A machine learning model was developed using three-year health examination data from 50,000 Japanese individuals.
• Heterogeneous Mixture Learning Technology (HMLT) was employed to generate predictive formulas.
• Model accuracy was validated on 5,000 individuals and compared against multiple regression using root mean square error (RMSE).

Main Results:

• The HMLT-based machine learning model generated five predictive formulas for body weight change.
• Lifestyle significantly influenced weight in individuals with high baseline BMI (≥29.93 kg/m²) and young individuals with low BMI (<23.44 kg/m²).
• The model achieved an RMSE of 1.914, comparable to multiple regression's RMSE of 1.890 (p=0.323).

Conclusions:

• The machine learning model effectively predicted three-year weight changes.
• The model identified specific demographic groups and lifestyle factors influencing weight dynamics.
• While requiring further validation across diverse populations, the model shows promise for individualized weight management in clinical settings.