Improved Heart Disease Diagnosis: Classification and Feature Selection Through Leveraging Machine Learning Algorithms

Abstract

Heart disease remains a leading cause of death globally, underscoring the urgent demand for precise and affordable diagnostic solutions. Given the complexity of predicting cardiac disease, advanced data mining techniques are essential for enhancing prediction accuracy. This work introduces a novel classification approach that identifies key predictive features, helping to reduce diagnostic costs without compromising accuracy.The methodology emphasizes robust data preprocessing, incorporating outlier detection through z-score analysis and effective feature selection, both of which significantly boost data quality and model reliability. Exploratory Data Analysis (EDA) is also instrumental in uncovering patterns and relationships between patient attributes and heart disease, supporting well-informed decisions during model development. The Random Forest classifier outperformed the other machine learning methods assessed, attaining an accuracy of 90.21%, which increased to 96.51% following feature selection. This outcome highlights the model’s capacity to handle complex datasets and underscores the importance of eliminating low-correlation features.The primary objective of this work is to enhance heart disease detection by integrating rigorous preprocessing techniques with powerful machine learning algorithms.