Performance Evaluation Of Ensemble Learning Using Light GBM For Enhanced Heart Disease Detection And Prediction

Abstract

Diseases of the heart (CVD) include the primary source of rising death rates as well as major cause of fatality. Improving the predictability as well as accuracy of cardiac disease is the primary goal of constructing the suggested model. Experts who ignore patient complaints put the patient at danger of serious complications that might result in death or disability. Consequently, in order to find patterns and hidden information in the medical data related to heart disease, we require expert systems that act as analytical tools. Finding hidden underlying patterns in vast amounts of data is a cognitive process known as machine learning. This study uses ensemble learning approaches in an attempt to improve the preciseness of the risk of heart disease assessment. Additionally, this research project has included feature selection and hyper parameter tuning approaches, which have increased accuracy even further. Used the information on heart disease to assess its performance using several measures. Six machine learning classifiers, including SVM, LR, RF, DT, and Ensemble techniques, were applied to the final dataset for this purpose, both before and after the hyper parameter tuning of the classifiers. Additionally, by doing specific data pre-processing, dataset standardization, and hyper parameter tweaking, using the common heart disease dataset, we confirm their correctness. The K-fold cross-validation approach was use through the researchers. Lastly, the experimental findings showed that machine learning classifiers' accuracy of prediction increased with hyper parameter tweaking, and they produced noteworthy outcomes with data standardization, hyper parameter tuning, and Light GBM.