Application of data mining techniques in predicting coronary heart disease: A systematic review

Abstract

Aim: The early detection of cardiovascular diseases by noninvasive and low-cost methods such as data mining techniques has been considered by many researchers. This study intends to review the studies performed on the prognosis of coronary heart disease using data mining techniques. Materials and Methods: The published studies in English between 2001 and 2021 that the use classification methods to predict coronary heart disease were considered. Databases such as ScienceDirect, Web of Science, and ScoPURs were considered as searchable databases. After searching, 348 articles were retrieved. After removing duplicates and evaluating the articles, finally, 20 articles were used. Results: The three data mining techniques support vector machine (SVM), neural network, and naive Bayes which were the most used among the studies. In the most studies, risk factors age, blood pressure, gender, diabetes, and chest pain were used. The accuracy was the most-used measure. The Alizadeh Sani dataset was the most used among the studies. Conclusion: Techniques such as SVM and neural network have performed better than other techniques. The output of these techniques can be used as a decision support system so that clinicians can enter various risk factors such as age, blood pressure, gender, diabetes, and chest pain and then view system output.