HYBRID FEATURE SELECTION STRATEGY FOR EFFICIENT HEART DISEASE CLASSIFICATION USING MACHINE LEARNING

The fact that heart disease continues to be one of the top causes of death across the globe calls for the implementation of decision support systems that are driven by data to facilitate early detection. To find the most discrimination features from a structured datasets for heart disease, this study suggests a hybrid feature selection methodology that combines filter and wrapper approaches. Prior to employing wrapper-based refinement with recursive feature reduction and classifier feedback, the method uses statistical correlation and mutual information ranking to eliminate superfluous features. Several machine learning classifiers, such as logistic regression, support vector machine, gradient boosting, and random forest, are trained using the optimum feature subset. Compared to baseline models trained on the whole set of features, the hybrid feature selection technique considerably improves classification accuracy, decreases dimensional, and boosts interpretations, according to experimental evaluation. Early diagnosis of cardiac illness with enhanced accuracy and computational efficiency is made possible by the methodology, which lays a solid groundwork for clinical decision support.