Performance of hybrid ensemble classification techniques for prevalence of heart disease prediction

Abstract

In medical science, heart disease is being considered as fatal problem and in every seconds most of the people dies due to this problem. In heart disease, typically heart stops blood supply to other parts of the body. Hence, proper functioning of body stopped and affected. In this way, timely and accurate prediction of heart disease is an important concern in medical science domain. Diagnosing of heart patients with previous medical history is not being considered as reliable in many aspects. However, machine learning techniques have mystery to classify heart disease data efficiently and effectively and provide reliable solutions. In the past, prediction of heart disease problem various machine learning tools and techniques have been adopted. In this study, hybrid ensemble classification techniques like bagging, boosting, Random Subspace Method (RSM) and Random Under Sampling (RUS) boost are proposed and performance is compared with simple base classification techniques like decision tree, logistic regression, Naive Bays, Support Vector Machine, k-Nearest Neighbor (KNN), Bays Net (BN) and Multi Layer Perceptron (MLP). The heart disease dataset from Kaggle data source containing 305 samples and Matlab R2017a machine learning tool are considered for performance evaluation. Finally, the experimental results stated that hybrid ensemble classification methods outperforms than simple base classification methods in terms of accuracy.