A hybrid cost-sensitive ensemble for heart disease prediction

Area of Science:

• Cardiology
• Machine Learning
• Medical Informatics

Background:

• Heart disease is a leading global cause of death and disability.
• Accurate heart disease diagnosis is challenging due to complex factors and high misclassification costs.

Purpose of the Study:

• To develop a cost-sensitive ensemble method for enhanced heart disease diagnosis.
• To improve diagnostic efficiency and minimize misclassification costs in cardiology.

Main Methods:

• An ensemble method combining five heterogeneous classifiers: random forest, logistic regression, support vector machine, extreme learning machine, and k-nearest neighbor.
• Utilized T-tests to evaluate ensemble performance against individual classifiers and assessed the impact of the Relief algorithm.

Main Results:

• The proposed ensemble method achieved superior performance via ten-fold cross-validation.
• Statistical tests confirmed the ensemble's significant superiority over individual classifiers.
• The Relief algorithm distinctly improved classification efficiency.

Conclusions:

• The developed ensemble method offers a promising alternative for medical decision-making in heart disease diagnosis.
• Demonstrated significantly better results compared to individual classifiers and prior studies.
• Highlights the potential of ensemble learning in improving cardiovascular disease diagnostic accuracy.