Machine Learning-Based Risk Factor Analysis and Prediction Model Construction for the Occurrence of Chronic Heart...

Area of Science:

• Cardiovascular Medicine
• Data Science
• Health Ecology

Background:

• Chronic heart failure (CHF) presents a significant global health challenge with high morbidity and mortality.
• Advancements in machine learning (ML) and health ecology offer new avenues for understanding CHF mechanisms and risk factors.
• Identifying high-risk individuals early is crucial for developing targeted prevention and intervention strategies.

Purpose of the Study:

• To develop an ML-based predictive model for CHF occurrence risk.
• To analyze CHF risk factors from a health ecology perspective.
• To compare the performance of various ML models for CHF risk prediction.

Main Methods:

• Utilized data from the Jackson Heart Study, including rigorous data preprocessing and feature selection using Principal Component Analysis and Random Forest (RF).
• Constructed and evaluated multiple ML models: Decision Tree, RF, Extreme Gradient Boosting, Adaptive Boosting (AdaBoost), Support Vector Machine, Naive Bayes, Multilayer Perceptron, and Bootstrap Forest.
• Validated model performance using 10-fold cross-validation and hyperparameter optimization, comparing metrics like accuracy, precision, sensitivity, F1-score, and AUC.

Main Results:

• RF outperformed PCA in feature selection, identifying 21 significant features.
• The AdaBoost model demonstrated superior performance with an initial AUC of 0.86, accuracy of 75.30%, precision of 0.86, sensitivity of 0.69, and F1-score of 0.76.
• Internal validation via 10-fold cross-validation yielded an AUC of 0.97 and accuracy of 91.27% for the AdaBoost model, further improved by hyperparameter optimization.

Conclusions:

• The study successfully developed and validated ML models for CHF risk prediction.
• AdaBoost emerged as the most effective model, highlighting the potential of ML in identifying individuals at risk of CHF.
• Future research should explore prospective studies, diverse datasets, advanced ML techniques, and longitudinal analyses to enhance CHF prevention and management.