An AI-Driven Automated Heart Disease Detection Model Using Heart Sounds: A Statistical and Machine Learning Approach with a Bagged Tree-Based Classifier

Heart disease detection is a critical global health challenge that demands accurate and advanced diagnostic approaches, particularly through the analysis of electrocardiogram (ECG) signals and heart murmurs. Early and precise identification of heart disease (HD) is essential, as it significantly reduces mortality rates through timely treatment. In recent years, substantial research efforts have been focused on developing intelligent and reliable systems for HD detection. This study proposes a machine learning-based framework for classifying heart disease using heart sound signals obtained from two well-established databases. To address class imbalance, the synthetic minority oversampling technique (SMOTE) was applied, followed by dimensionality reduction using principal component analysis (PCA). For the classification task, a bagged tree-based classifier (BTBC) has been employed, achieving a high accuracy of 99.03% compared to various traditional models. Additionally, detailed statistical tests like Pearson’s correlation, t-tests, ANOVA, and Chi-Square tests have been performed to find important relationships and differences in the data, which strengthens the reliability of the suggested method.