An Optimized Stacked Support Vector Machines Based Expert System for the Effective Prediction of Heart Failure

• Published in: IEEE access Vol. 7; pp. 54007 - 54014
• Main Authors: Ali, Liaqat, Niamat, Awais, Khan, Javed Ali, Golilarz, Noorbakhsh Amiri, Xingzhong, Xiong, Noor, Adeeb, Nour, Redhwan, Bukhari, Syed Ahmad Chan
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 2019; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Clinical expert system; Correlation coefficients; Diagnosis; Diseases; Expert systems; feature selection; Heart; Heart failure; heart failure prediction; hybrid grid search algorithm; Kernel; Machine learning; Optimization; Performance enhancement; Prediction models; Predictive models; Search algorithms; support vector machine; Support vector machines
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2019.2909969

About half of the people who develop heart failure (HF) die within five years of diagnosis. Over the years, researchers have developed several machine learning-based models for the early prediction of HF and to help cardiologists to improve the diagnosis process. In this paper, we introduce an expert system that stacks two support vector machine (SVM) models for the effective prediction of HF. The first SVM model is linear and <inline-formula> <tex-math notation="LaTeX">L_{1} </tex-math></inline-formula> regularized. It has the capability to eliminate irrelevant features by shrinking their coefficients to zero. The second SVM model is <inline-formula> <tex-math notation="LaTeX">L_{2} </tex-math></inline-formula> regularized. It is used as a predictive model. To optimize the two models, we propose a hybrid grid search algorithm (HGSA) that is capable of optimizing the two models simultaneously. The effectiveness of the proposed method is evaluated using six different evaluation metrics: accuracy, sensitivity, specificity, the Matthews correlation coefficient (MCC), ROC charts, and area under the curve (AUC). The experimental results confirm that the proposed method improves the performance of a conventional SVM model by 3.3%. Moreover, the proposed method shows better performance compared to the ten previously proposed methods that achieved accuracies in the range of 57.85%-91.83%. In addition, the proposed method also shows better performance than the other state-of-the-art machine learning ensemble models.