A decision support system for heart disease prediction based upon machine learning

Detection of heart disease through early-stage symptoms is a great challenge in the current world scenario. If not diagnosed timely then this may become the cause of death. In developing countries where heart specialist doctors are not available in remote, semi-urban, and rural areas; an accurate decision support system can play a vital role in early-stage detection of heart disease. In this paper, the authors have proposed a hybrid decision support system that can assist in the early detection of heart disease based on the clinical parameters of the patient. Authors have used multivariate imputation by chained equations algorithm to handle the missing values. A hybridized feature selection algorithm combining the Genetic Algorithm (GA) and recursive feature elimination has been used for the selection of suitable features from the available dataset. Further for pre-processing of data, SMOTE (Synthetic Minority Oversampling Technique) and standard scalar methods have been used. In the last step of the development of the proposed hybrid system, authors have used support vector machine, naive bayes, logistic regression, random forest, and adaboost classifiers. It has been found that the system has given the most accurate results with random forest classifier. The proposed hybrid system was tested in the simulation environment developed using Python. It was tested on the Cleveland heart disease dataset available at UCI (University of California, Irvine) machine learning repository. It has achieved an accuracy of 86.6%, which is superior to some of the existing heart disease prediction systems found in the literature. © 2021, The Author(s), under exclusive licence to Springer Nature Switzerland AG part of Springer Nature.