期刊名称：Indian Journal of Computer Science and Engineering
印刷版ISSN：2231-3850
电子版ISSN：0976-5166
出版年度：2021
卷号：12
期号：4
页码：1001-1017
DOI：10.21817/indjcse/2021/v12i4/211204182
语种：English
出版社：Engg Journals Publications
摘要：Heart diseases cause most of the untimely and sudden deaths all over the world. Many of the lives can be saved by an early alarm with the means of expert diagnosis systems. Researchers developed such expert systems over time. However, there is still a need for more efficient methods for accurate and efficient diagnosis of heart disease. The accuracy and efficiency of the diagnosis system are highly dependent on the characteristics of data, feature selection (FS) algorithms and classification techniques. The artificial neural network (ANN) exhibits excellent performance on unseen data samples and metaheuristic optimization techniques have performed exceptionally well in training the ANN model and avoiding local minima problems. In this article, the authors have employed an artificial neural network (ANN), trained with a Grey Wolf Optimizer (GWO) and Particle Swarm Optimization (PSO) empowered, Inertia Motivated GWO (IMGWO) technique, to classify a real-world dataset. Three most popular feature selection techniques viz. Relief, mRMR and LASSO are employed to find out the most discriminative features from the collected dataset. After that, the selected features are utilized in the IMGWO trained ANN model for classification. The classification performance of the said technique is then compared with the results obtained from five most prevalent metaheuristic methods viz. Genetic Algorithm (GA), Firefly Algorithm (FF), Whale Optimization Algorithm (WOA), PSO and GWO used to train the ANN model. It is observed from the classification results that the performance of the IMGWO technique utilizing the mRMR feature selection algorithm supersedes all other techniques used. The results have been compared in terms of performance metrics that include accuracy, sensitivity, specificity, the area under the curve (AUC), F-measure, precision, kappa statistics and Mathew’s correlation coefficients. The work will be a contribution towards the development of expert systems for heart disease diagnosis with an improved efficacy.
