A Novel Heart Disease Prediction System Using XGBoost Classifier Coupled With ADASYN SMOTE

According to official data from the World Health Organization, cardiovascular disease stands as the supreme cause of global fatalities. The estimated number is 17.9 million deaths [16] per year. To provide cure proper recognition of illness based on prevailing symptoms is required. Traditional techniques face many problems like delayed or unnecessary treatment to misdiagnosis. This results into increasing health problems of patient and a big financial loss. The situation can be improved by using AI and ML techniques. Over the past few years, substantial endeavors have been dedicated to enhance applications related to computer-aided diagnosis and detection. Previous researches are based on just medical records of patients and not on their day to day lifestyle habits. In this study we have proposed a prediction system using Extreme Gradient Boost (XGBoost) Classifier coupled with ADASYN SMOTE (used to balance the dataset) that can predict the possibility of getting heart disease based on some observations related to lifestyle. The model has achieved the accuracy of 94.7
^%
on the dataset Personal Key Indicators of Heart Disease that is based on a recent data collected by Centre Of Disease Control and Prevention (https://www.kaggle.com/datasets/kamilpytlak/personal-key-indicators-of-heart-disease). The dataset includes the information of around four lakhs people who have shared some basic information related to their health including the fact that they have experienced Coronary Heart Disease (CHD) or Myocardial Infarction (MI). The accuracy of 94.7% has been achieved with XGBoost classifier (highest among six classifiers) on a balanced data which means model will not give biased results and also it is better than previous work [23]. As the dataset is not balanced we have proposed the use of ADASYN SMOTE over SMOTE to balance the dataset.