Research Article

A comparative study of soft computing models for heart disease classification

Authors

  • Victoria Nozick * Operations and Information Management Group, Aston University, B4 7ET Birmingham, United Kingdom.

https://doi.org/10.22105/thi.v3i3.67

Abstract

 

Accurate classification of patients with heart disease remains an important challenge in clinical decision support, particularly when several clinical variables must be considered simultaneously. Computational methods, including artificial neural networks and other soft computing techniques, offer a way to identify patterns in clinical data and support the classification process. This study evaluates and compares seven computational models for heart disease classification, with particular attention to the performance of a Generalized Regression Neural Network (GRNN). The study used a dataset containing 350 patient records. The classification process was based on clinical attributes describing patients' demographic characteristics, medical history, and examination results. The available records were divided into training, validation, and testing subsets, and 52 records were used for testing. Seven models, namely GRNN, Multilayer Perceptron (MLP), Probabilistic Neural Network (PNN), Self-Organizing Map (SOM), Support Vector Machine (SVM), Time-Delay Neural Network (TDNN), and Generalized Feedforward Network (GFF), were implemented and evaluated using NeuroSolutions. The results showed that GRNN achieved the highest overall classification accuracy, correctly classifying 94.23% of the test records. PNN ranked second with an accuracy of 88.46%, followed by MLP with 65.38%. The corresponding accuracies for SOM, TDNN, GFF, and SVM were 53.85%, 55.77%, 51.92%, and 42.31%, respectively. GRNN also achieved an (R^2) value of 0.960, a Mean Absolute Error (MAE) of 0.093, a Normalized Root Mean Square Error (NRMSE) of 0.061, and a Root Mean Square Error (RMSE) of 0.246. Its class-specific results were also comparatively strong, although performance varied across the four output classes. Overall, the comparative results indicate that GRNN provided the most accurate classification among the seven evaluated models under the experimental conditions of this study. These findings support further investigation of GRNN-based approaches for clinical decision-support applications. However, the results should be interpreted in light of the dataset size and the evaluation procedure reported in the study.        

Keywords

Heart disease classification Soft computing Generalized regression neural network Artificial neural networks Clinical decision support Classification accuracy
References (21)
  1. [1] Chew, D. S., Cowper, P. A., Al-Khalidi, H., Anstrom, K. J., Daniels, M. R., Davidson-Ray, L., ... & STICH Investigators. (2022). Cost-effectiveness of coronary artery bypass surgery versus medicine in ischemic cardiomyopathy: The STICH randomized clinical trial. Circulation, 145(11), 819–828. https://doi.org/10.1161/CIRCULATIONAHA.121.056276

  2. [2] Wilson, M. P. W., Castelli, M. W. P., & Kannel, M. W. B. (1987). Coronary risk prediction in adults (the Framingham Heart Study). The American Journal of Cardiology, 59(14), G91–G94. https://doi.org/10.1016/0002-9149(87)90165-2

  3. [3] Baxt, M. W. (1995). Application of artificial neural networks to clinical medicine. The Lancet, 346(8983), 1135–1138. https://doi.org/10.1016/S0140-6736(95)91804-3

  4. [4] Jain, A. K., Mao, J., & Mohiuddin, K. M. (1996). Artificial neural networks: A tutorial. Computer, 29(3), 31–44. https://doi.org/10.1109/2.485891

  5. [5] Palaniappan, S., & Awang, R. (2008). Intelligent heart disease prediction system using data mining techniques. In 2008 IEEE/acs international conference on computer systems and applications (pp. 108–115). IEEE. https://doi.org/10.1109/AICCSA.2008.4493524

Article Metrics

How to Cite

Nozick, V. (2026). A comparative study of soft computing models for heart disease classification. Trends in Health Informatics, 3(3). https://doi.org/10.22105/thi.v3i3.67

Similar Articles

1-10 of 37

You may also start an advanced similarity search for this article.