Advanced Ensemble Machine Learning Approach for ECG-Based Arrhythmia Detection

Cardiovascular diseases (CVDs) are the main reason causing millions of deaths around the world. To limit the negative effects of CVDs, early arrhythmia detection is important. The electrocardiogram (ECG) is one of the most popular methods to monitor the status of the heart helping to detect irregular heartbeats easily and supporting both doctors and general users in the clinical stage. In this paper, the ability of ensemble Machine Learning (ML) algorithms and 5-fold cross-validation (CV) combined with multiple effective signal processing techniques such as median and Butterworth (BW) filters, Ensemble Empirical mode decomposition (EEMD) procedure and Hilbert transform (HT) is utilized to propose an automatical arrhythmia detection to classify between two classes: Regular HeartBeat and Irregular HeartBeat. The given results are outstanding with 96.43% accuracy (Acc) and 98.95% Area Under Curve score (AUC), which can become a potential clinical method to contribute to health facilities, especially in developing countries.