Abnormal ECG Detection using Optimized Boosting Tree Classifier

Abstract

ECG plays an important role in cardiac disease diagnosis. Classification of this cardiac signal using machine learning techniques will be a supportive tool for the physicians. Authors in this work have classified the ECG by using three different types of classifiers such as Support vector machine (SVM), Gradient boosting, and extreme gradient boosting (XGBoost). The standard statistical features are considered as input to the classifiers. For improving the learning strategy and performance of the proposed models subjected to accuracy, the learning rates are varied for each node of the tree-based ensemble classifiers. Also, the hyperparameters of the XGBoost model are optimized by applying the Bayesian optimization (BO) technique. The best accuracy in SVM classifier is found as 91.69%. 96.58% accuracy is obtained in the modified gradient boosting model. The optimized XGBoost model is providing 100% accuracy which is better than other.