Machine Learning-Based Ensemble Predictive Model for Cardiovascular Disease Prevention.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Medicine
• Public Health

Background:

• Cardiovascular diseases (CVDs) represent a leading global cause of mortality, with a notable increase in incidence within India.
• Current machine learning (ML) models for CVD prediction face limitations due to dataset size and generalizability, impacting their robustness.

Purpose of the Study:

• To develop an enhanced ML prediction model for CVD detection utilizing ensemble methods.
• To improve the accuracy and robustness of CVD prediction models by leveraging extensive clinical datasets.

Main Methods:

• Consolidated six datasets comprising 7,916 clinical records, divided into Dataset 1 (n=3,676) and Dataset 2 (n=4,240) based on features.
• Applied binary and multiclass classification for Dataset 1, and binary classification (10-year risk) for Dataset 2, following identical preprocessing and exploratory data analysis.
• Developed 18 distinct ML models, selected the top 10 performers using LazyPredict, and integrated them into an ensemble model via Voting Classifier.

Main Results:

• Achieved 96.5% accuracy for binary classification and 85.5% for multiclass classification on Dataset 1.
• Attained 81.18% accuracy for Dataset 2's 10-year CVD risk prediction.
• The ensemble framework demonstrated superior performance compared to traditional and existing ML models.

Conclusions:

• The proposed ensemble ML framework significantly enhances CVD prediction accuracy and stability.
• The study provides improved decision support for CVD detection, mitigating bias through robust modeling.
• This approach addresses limitations of dataset size and generalizability in current ML-based CVD prediction systems.