Comparative Performance Analysis of Decision Tree Variants in Heart and Lung Disease Classification

Machine learning (ML) algorithms are pivotal for disease detection, with decision tree (DT) models particularly prominent in classification tasks. This paper conducts a thorough comparative analysis of different variants of DT algorithm for heart and lung disease classification. DT variants, such as ID3, C4.5, CART, and CHAID were evaluated on medical image datasets, addressing missing values and class imbalances through preprocessing. Accuracy, precision, recall, and F1-score were employed for assessment, and CHAID exhibited superior accuracy and robustness across both diseases. CART and C4.5 were competitive but fell short in predictive power. The study emphasizes selecting the right DT variant based on diagnostic requirements and explores ensemble methods’ potential to enhance classification outcomes. These findings offer valuable insights for developing more effective decision-support systems in clinical settings, thereby improving disease diagnosis accuracy and reliability.