A Comparative Study of Machine Learning Techniques in Heart Disease Detection

Cardiovascular diseases (CVD) are the world's leading cause of mortality accounting for an estimated 31% of all deaths worldwide. Out of 17.9 million deaths per year due to CVDs, three-fourths of these deaths have occurred as there are no systems in place to predict the occurrence of a heart attack and warn the patient or doctor to take appropriate action. Data generated by clinical reports and examination reports by doctors are available for prediction through ERP models. Data science and reliable algorithms powered by AI can be used to develop medical devices that can predict such incidents of CVDs. In this paper, seven common classifiers are implemented that are computationally inexpensive and easily implementable and their performance metrics are compared. Two feature selection techniques are implemented and Grid Search is used for hyper-parameter tuning. Using k-fold cross-validation, classifiers are then evaluated, which generates classification metrics such as accuracy, f1-score, recall, and precision. It is evident from the study that the combination of Random Forest Classifier and SelectKBest feature selector has the highest accuracy of 89.706% and precision of 89.655%.To view and download the paper for free, visit: http://pices-journal.com/ojs/index.php/pices/article/view/28