Myocardial ınfarction classification with support vector machine models

Abstract

Aim: Support vector machines (SVM) is one of the classification methods that aims to find the best hyper-plane separating a space into two parts with known positive and negative samples. The goal of this study is to classify myocardial infarction (MI) using SVM models. Material and Methods: The data used in the MI classification contains information related to 184 individuals which is randomly taken from the database created for the Department of Cardiology, Faculty of Medicine, Inonu University. Estimated SVMs are models generated from the SVM-linear and SVM-Radial Based kernel functions. Results: In this study, 90 individuals of the study group (48.9%) are MI patients, while 94 (51.1%) patients are not. The classification success rate is 83.70% for SVM-linear model and 90.76% for the SVM-Radial Based model. Conclusion: In this study, it is observed that SVM-Radial based model presented a better classification performance than the linear SVM model. The use of SVM models based on various kernel type functions can improve disease classification performance.