A Novel Prediction System to Diagnose Heart Disease

Worldwide, heart disease is among the leading killers. This illness accounts for around 31 ^% of all fatalities worldwide. The World Health Organization estimates that over 17.9 million individuals each year lose their lives to this illness. If heart disease is diagnosed at an early stage and treated, the patient has a better chance of avoiding the illness's more severe later stages. While coronary angiography provides the most accurate detection of the condition, individuals often avoid it in the early stages due to its discomfort, expense, and invasive nature. Thus, there is a need for a non-invasive and reliable approach that may be used as a screening tool for illness diagnosis. Every aspect of modern life has been impacted by machine learning, and the health care industry is no exception. The clinical characteristics of a patient can be evaluated using machine learning based decision support systems to help in the diagnosis of cardiac disease. With the use of clinical criteria, a novel prediction system for cardiac disease has been developed. Although the MICE approach was utilized to deal with missing values, the hybrid GARFE algorithm was used for feature selection and further the SMOTE (synthetic minority oversampling technique) and the usual scalar techniques were used for pre-processing the data. In the end, a variety of classifiers were applied, including support vector machines, naive bayes, logistic regression, random forest, and adaptive boosting classifiers. Based on the data, it appears that the proposed system has achieved an accuracy of 86.60% using Random Forest (RF) with the Cleveland dataset.