Advanced Ensemble Machine Learning for ECG Arrhythmia Classification

Early diagnosis and classification of cardiac arrhyth-mias from electrocardiogram (ECG) recordings play a crucial role in smart healthcare systems for monitoring the health of individuals with heart conditions. However, the nonlinear nature and low amplitude of ECG recordings pose challenges to accurate classification, and the performance of traditional machine learning (ML) classifiers remains uncertain due to poorly modeled correlations between learning parameters. Ad-dressing these limitations, this study introduces an automated cardiac arrhythmia classification framework, leveraging effective signal processing techniques including median filtering, Butter-worth (BW) filtering, Ensemble Empirical Mode Decomposition (EEMD), and Hilbert Transform (HT). Subsequently, optimized input parameters for ML classifiers are determined using the Marine Predator Algorithm (MPA) on the training set, coupled with cross-validation (CV) to identify optimal parameters for each model. Finally, the proposed Random Forest (RF) model demonstrates high classification performance on the test dataset with optimized input hyperparameters, achieving an average accuracy of 99.90%, precision of 99.74%, recall of 99.73%, and Fl score of 99.74%. This proposed method outperforms in terms of accuracy across all classes, achieving 99.74 %, 99.92 %, 99.89 %, 99.98%, and 99.94% for classes normal (N), supraventricular ectopic (S), ventricular ectopic (V), fusion (F), and unknown (Q), respectively. These results highlight the potential of this approach as a robust clinical tool, particularly in developing countries.