Comparing the performances of support vector machines, artificial neural network, and a logistic regression model for predicting coronary artery diseases: a cross‑sectional study

Abstract

Abstract Background Coronary artery disease (CAD) is considered as an inflammatory disease. Cardiovascular disease (CVD) is a major cause of death and disability worldwide. This study aimed to compare the performance of different non-invasive CAD diagnostic techniques. Methods A cross-sectional study was performed on a total of 758 subjects (250 with CAD and 508 without CAD). We compared the performances of logistic regression (LR) model, artificial neural networks (ANN), and support vector machines (SVMs) for the purpose of functioning. The Performance of classification techniques were compared using ROC curve, sensitivity, specificity, and accuracy. Results The study population consisted of 758 case subjects. Two hundred fifty of them (33.6% men and 66.4% women) were eventually diagnosed with non-CAD, while 508 subjects (64% men and 36% women) were not (33.6% men and 66.4% women). The area under the ROC Curve (AUC) for CAD resulted in 0.775 (95% CI: 0.711, 0.838) for Logistic regression model, 0.752 (95% CI: 0.682, 0.823) for ANN, and 0.793 (95% CI: 0.733, 0.853) for SVMs, respectively. There were significant differences between these three models in prediction of CAD (p = 0.04). The best model of forecasting CAD was the SVMs (0.793, 95% CI: 0.733, 0.853). However, the differences between logistic regression model, ANN and LR with SVMs were small and non-significant (p = 0.2, p = 0.09). Conclusions Support vector machines (SVMs) yielded better performance than ANN model to predict the risk of coronary artery disease (CAD) with simple clinical predictors. However, support vector machines produced as much performance as the LR model.