Feature Integration and Decision Tree Optimization for Improved Coronary Artery Disease Classification

Heart disease, specifically coronary artery disease (CAD), is one of the biggest health issues globally, which calls for accurate and reliable methods in diagnosis. This study utilizes data from several CAD-related sources: Cleveland, V.A., Hungarian, and Statlog datasets, with the goal of maximizing accuracy using machine learning techniques. The applied techniques include Fast Decision Tree and pruned C4.5 decision trees on individual sources and on the combined database with key features obtained using entropy-based methods. Other models, such as Logistic Regression, k-Nearest Neighbors, Random Forest, and Neural Networks, were also tested on the integrated data. By merging datasets, this paper addresses common challenges like missing or inconsistent data, thus achieving better performance. In particular, the Random Forest model delivered the best results with a 91.75% accuracy rate. This approach demonstrates how merging datasets and utilizing advanced algorithms may lead to more reliable diagnoses while also pointing out the differences that can arise when analyzing separate data sources.