Early Prognosis of Coronary Heart Disease using Ensemble Classifiers: A Comparative Analysis

Abstract

Machine Learning techniques are extensively used in health care especially for disease prediction.  This paper presents a comparison of performance of various Ensemble classifiers for early detection of Coronary Heart Disease (CHD) based on risk factors.  This paper focuses on the Bagging, Boosting and Subspace Ensemble Classifiers for detecting CHD.  The performance of these ensemble classifiers is compared with performance indices such as accuracy, precision, recall and F1 score. K-Fold’s validation is adopted to randomize the data and to obtain the consistency of results. In the current research work, the experimentation has been carried out on the datasets acquired from UCI dataset.  From the experimentation results, it is observed that Bagged Trees Ensemble classifier provides a highest classification accuracy, precision, recall and F1 score of 95.5 %, 0.95, 0.97 and 0.95 respectively for identifying CHD. The result also depicted that the Bagged Trees Ensemble classifier outperformed in comparison with the traditional classifiers. The current work is useful for physicians to detect the coronary heart disease at early stages.