Using Ensemble Approaches and Different Sampling Techniques to Handle Class Imbalance Challenges in Coronary Heart Disease Prediction

Coronary Heart Disease (CHD) has a significant impact on death rates worldwide; therefore, early and accurate prediction is vital for enhancing timely prevention and improving patient outcomes. However, predicting CHD is challenging because of the imbalanced nature of the dataset. A number of sampling techniques are proposed to overcome this challenge, including random oversampling, undersampling, and synthetic minority oversampling techniques (SMOTE), as well as three ensemble methods, including XGBoost, Random Forest, and AdaBoost. Incorporating these techniques minimizes classifier bias in favor of the majority group and emphasizes the learning of minority concepts appropriately. Undersampling techniques give the highest accuracy in AdaBoost, which is 53%. However, the oversampling approach and SMOTE achieved the highest accuracy of 82%, which is in the Random Forest algorithm. The sensitivity and precision scores are also increased in the proposed model. Therefore, our proposed approach offers a trustworthy and effective solution to the class imbalance issue in CHD prediction, and it has the potential to be used in other healthcare areas that face similar challenges.

I. Introduction

Coronary heart disease (CHD) is one of the significant causes of disability and mortality among cardiovascular diseases (CVDs), which is also the leading cause of death globally. A total of 17.9 million people died from CVDs in 2019, accounting for 32% of all deaths worldwide. The cause of 85 % of these deaths was heart attacks and strokes [1]. Chest pain or discomfort (angina), shortness of breath, or even a heart attack can result from CHD when the blood arteries supplying the heart with blood and oxygen narrow or get clogged. Age, gender, family history, smoking, obesity, high cholesterol, diabetes, hypertension, and hyperlipidemia are some of the numerous risk factors that might contribute to this complicated disease. Finding patients who have a high risk of getting CHD is a crucial job in CHD prediction. It is vital for the disease's early identification, treatment, and control [2].