Diagnosis of coronary artery disease based on machine learning algorithms support vector machine, artificial neural network, and random forest
Saeed Saeedbakhsh1, Mohammad Sattari1, Maryam Mohammadi2, Jamshid Najafian3, Farzaneh Mohammadi4
1 Health Information Technology Research Center, Isfahan University of Medical Sciences, Isfahan, Iran
2 Department of Management and Health Information Technology, School of Management and Medical Information Sciences, Isfahan University of Medical Sciences, Isfahan, Iran
3 Isfahan Cardiovascular Research Center, Cardiovascular Research Institute, Isfahan University of Medical Sciences, Isfahan, Iran
4 Department of Environmental Health Engineering, School of Health, Isfahan University of Medical Sciences, Isfahan, Iran
Ms. Maryam Mohammadi
School of Medical Management and Information Sciences, Isfahan University of Medical Sciences, Hezarjerib Avenue, P. O. Box: 81745-346, Isfahan
Dr. Mohammad Sattari
Department of Health Information Technology and Management, School of Medical Management and Information Sciences, Isfahan University of Medical Sciences, Hezarjerib Avenue, P. O. Box: 81745-346, Isfahan
Source of Support: None, Conflict of Interest: None
Background: Coronary artery disease (CAD) is known as the most common cardiovascular disease. The development of CAD is influenced by several risk factors. Diagnostic and therapeutic methods of this disease have many and costly side effects. Therefore, researchers are looking for cost-effective and accurate methods to diagnose this disease. Machine learning algorithms can help specialists diagnose the disease early. The aim of this study is to detect CAD using machine learning algorithms.
Materials and Methods: In this study, three data mining algorithms support vector machine (SVM), artificial neural network (ANN), and random forest were used to predict CAD using the Isfahan Cohort Study dataset of Isfahan Cardiovascular Research Center. 19 features with 11495 records from this dataset were used for this research.
Results: All three algorithms achieved relatively close results. However, the SVM had the highest accuracy compared to the other techniques. The accuracy was calculated as 89.73% for SVM. The ANN algorithm also obtained the high area under the curve, sensitivity and accuracy and provided acceptable performance. Age, sex, Sleep satisfaction, history of stroke, history of palpitations, and history of heart disease were most correlated with target class. Eleven rules were also extracted from this dataset with high confidence and support.
Conclusion: In this study, it was shown that machine learning algorithms can be used with high accuracy to detect CAD. Thus, it allows physicians to perform timely preventive treatment in patients with CAD.