Predicting risk of obesity in overweight adults using interpretable machine learning algorithms

Area of Science:

• Machine Learning
• Predictive Analytics
• Public Health

Background:

• Obesity is a growing public health concern globally.
• Identifying predictive factors for obesity is crucial for effective prevention and management strategies.
• Machine learning offers powerful tools for analyzing complex health data to identify risk factors.

Purpose of the Study:

• To screen for predictive obesity factors in overweight populations.
• To identify an optimal and interpretable machine learning algorithm for obesity risk prediction.
• To utilize advanced machine learning techniques for public health insights.

Main Methods:

• A cross-sectional study involving 5,236 Chinese participants.
• Seven machine learning methods were employed to build obesity risk prediction models.
• The CatBoost algorithm was selected as the best-performing model, validated using AUC and cross-validation.
• Shapley additive explanation was used for model interpretability.

Main Results:

• CatBoost demonstrated strong performance in predicting obesity, with AUC values of 0.95 (training) and 0.87 (test).
• Key predictors identified include waist circumference, hip circumference, female gender, and systolic blood pressure.
• The model's validity and net clinical benefit were superior compared to other methods.

Conclusions:

• CatBoost is a highly effective machine learning method for obesity risk prediction.
• Integrating machine learning with Shapley additive explanation aids in identifying and understanding disease risk factors.
• These findings can inform targeted prevention and control strategies for obesity.