Cardiovascular disease is the general name for many disorders relevant to heart diseases. There are a
particularly coronary artery disease (CAD) (Libby et al. 2011). CAD causes a reduced flow of blood to the heart muscles (“Global atlas on cardiovascular disease prevention and control / edited by: Shanthi Mendis ... [et al.]” n.d.). This disease is known as important clinical disorders that occur due to the abnormal functioning of the heart (Ravi et al. 2017). The blockage or narrowing of blood vessels in the heart causes a heart disease or stroke (Prabhakaran et al. 2016). Cardiovascular diseases, which include heart disease, cerebrovascular disease, and blood vessel disease, are one of the leading causes of death worldwide (“Global atlas on cardiovascular disease prevention and control / edited by: Shanthi Mendis ... [et al.]” n.d.). A good deal of people is affected by heart disease, especially CAD. World Health Organization reported that approximately 17.9 million people die each year and it is estimated that 32% of the deaths in the world are due to this disease (WHO, 2022). CAD-related deaths could be prevented by accurate detection and timely intervention (Abdar et al. 2019).
Early diagnosis and appropriate treatment in heart disease can reduce and prevent the death rate of patients. Angiography is a common method of diagnosing abnormal narrowing of the heart vessel (Alizadehsani, Habibi, Sani, et al. 2013). Accurate and early diagnosis of heart disease allows timely and accurate treatment and also reduces mortality (Ashish et al. 2021). Moreover, with timely treatment, the severity of the side effects of CAD can be reduced. Angiography is used to detect stenosis and its location in the heart vessels (Alizadehsani, Habibi, Hosseini, et al. 2013). In order to diagnose CAD, field experts use different methods based on angiography, which is considered a definitive solution (Arabasadi et al. 2017). Coronary angiography (Alizadehsani et al. 2019) is costly and time-consuming. Therefore, the machine learning-based systems in CAD diagnosis will help field experts.
In literature, it is seen that different machine learning aproaches have been used for the CAD problem, so far. However, most of these studies were carried out on outdated datasets such as Cleveland, Hungarian. For example, Polat et al. firstly obtained new values for each attribute based on the k-nearest neighbor method and the authors performed an artificial immune recognition system with fuzzy resource allocation mechanism based classification to detect heart disease (Polat et al. 2007). In another study, Bahani et al. proposed a method based on a fuzzy rule-based classification (Bahani et al. 2021). Gárate-Escamila et al. performed experiments with 6 classifiers on the features which are the combination of chi-square and principal component analysis on the heart disease dataset downloaded from the UCI Machine Learning Repository (Gárate-Escamila et al. 2020). Lastly, Jabbar et al. proposed a genetic algorithm and k-Nearest Neighbors based approach to improve accuracy for diagnosing heart disease (jabbar et al. 2013).
In recent years, the studies based on artificial intelligence and machine learning have been very popular. Different solutions are obtained with different models by performing machine learning studies on the data obtained from patients (Condie et al. 2013). Machine learning-based studies are often used to investigate the relationship between attributes and target classes.
This study was conducted with the motivation of applying machine learning techniques on CAD data in order to develop decision supporting systems that give support specialist doctors in their fields. The main purpose of this study is to identify the most successful model for a decision support system that detects CAD at an earlier stage with minimum error. In line with this, a comprehensive comparison of the models based on three scnerios was made. The main contributions of this study are as follows:
- This study investigates a prospering machine learning method in detecting CAD and proposes a machine learning model that detects this disease with good accuracy.
- A hybrid model, which includes SVM optimized with Randomize Search cross validation that has not been used in previous studies, is proposed to classify CAD.
- The performances of the models are validated on the Z-Alizadeh Sani heart disease dataset with the 5-fold cross-validation techniques.
The remainder of the study is organized as follows: Section 2 provides detailed information on materials and methods. Section 3 introduces the experimental procedure within the data preparation and data classification steps. Section 4 presents the experimental results in detail and discusses with the studies in literature. Finally, Section 5 concludes the work with final remarks.