An Efficient and Explainable Model for Heart Disease Classification using Boosting Algorithms

Heart diseases (HDs) contribute to about 32% of deaths worldwide, according to the World Health Organization (WHO). Individuals experiencing high blood pressure, high cholesterol, and other risk factors are more susceptible to heart disease. This creates a need for a sophisticated and explainable system to classify heart disease patients based on a range of clinical data. This paper proposes an efficient and explainable boosting-based framework to classify HD patients. Data balancing is performed on the dataset based on classification labels. Boosting models are used to classify HD patients. The classification performance of each model is measured using various quantitative metrics with values over 93% demonstrating excellent classification performance. SHAP is integrated to explain the HB model and assess the importance of each feature. Upon explanation, the number of major blood vessels, chest pain type, thalassemia, and ST depression were found to be the most impactful features for HD classification. The combination of strong quantitative performance and explainable AI makes our method suitable for use by medical practitioners for quick and automated diagnosis of HD patients.