A Robust MI-Based Hybrid Diagnostic Model for Early Detection of Heart Diseases

Abstract

Heart disease operates as one of the leading dangerous causes of death worldwide thus humans require both precise and speedy medical diagnosis applications.   Machine learning (ML) exhibits impressive potential to boost clinical decision-making each year because it effectively duplicates patterns within complex HiMed data. Machine learning demonstrates pattern imitiation through this capability.   The main purpose of this research involved the development of a hybrid machine learning system that predicted heart disease. The system utilizes majority voting ensemble method to unite SVM with DT and RF classifiers for prediction purposes.   The research utilizes the Cleveland Heart Disease dataset found at UCI Machine Learning Repository to conduct training and testing operations.   The preprocessing procedures contain One-hot category encoding together with normalization of data and Recursive Feature Elimination (RFE) feature selection functionality.   The suggested hybrid combination model achieves 92.5% accuracy and 91.8% precision while reaching 93.2% recall and 92.5% F1-score making it perform better than single classifiers.   The findings match with the conclusion about the hybrid ensemble approach being more resilient with general capabilities and diagnostic accuracy. Such systems prove to be an excellent practical solution for operational medical decision programs used in actual healthcare settings.