Comparative Analysis of Machine Learning Algorithms for Arrhythmia Detection

The presented paper proposes a framework for arrhythmia classification ECG signal. The proposed framework is validated using the MIT-BIH Arrhythmia Database. The class imbalance is mitigated with the SMOTE technique. The performance of five different classifiers is evaluated and compared using accuracy, precision, recall, and F1 score. Results show that Random Forest achieves 98.32% accuracy, 89.12% precision, 92.29% recall, and 90.57% F1-score. Whereas the decision tree emerges as the 2nd best classifier. The obtained results highlight the efficacy of the proposed framework for automated ECG analysis in clinical diagnostics.