Machine learning-based diagnosis and risk factor analysis of cardiocerebrovascular disease based on KNHANES

Area of Science:

• Computational epidemiology
• Machine learning in healthcare
• Public health surveillance

Background:

• Cardiocerebrovascular disease (CVD) prevalence is rising globally, posing a significant public health challenge.
• Accurate and interpretable screening methods are crucial for early detection and prevention of CVD.
• Existing screening methods face limitations in scalability and precision for large populations.

Purpose of the Study:

• To develop and evaluate machine learning classifiers for predicting CVD risk.
• To identify key risk factors contributing to CVD prevalence using interpretable AI.
• To enhance the accuracy and interpretability of CVD screening tools.

Main Methods:

• Utilized Korea National Health and Nutrition Examination Survey (KNHANES) data.
• Applied advanced machine learning algorithms: multi-layer perceptron, support vector machine, random forest, and light gradient boosting.
• Employed data rebalancing techniques (SMOTE, random undersampling) and feature selection (VIF, Boruta) for improved model performance and interpretability.

Main Results:

• Achieved excellent classifier performance with Area Under the Curve (AUC) values exceeding 0.853.
• Identified age, sex, and hypertension as the most significant risk factors for CVD.
• Revealed positive correlations between CVD and age, hypertension, and BMI; negative correlations with female sex, alcohol consumption, and higher income.

Conclusions:

• Machine learning models offer a highly accurate approach to CVD risk prediction.
• Feature selection and class balancing techniques significantly enhance model interpretability.
• The findings provide valuable insights for targeted public health interventions and CVD prevention strategies.