Ensemble Learning Methodologies for Electrocardiogram Analysis: A Comparative Study

Electrocardiogram (ECG) analysis plays a crucial role in diagnosing various cardiovascular diseases. In this study, we present a comprehensive analysis of ensembled machine learning methodologies for ECG classification. Specifically, we evaluate the performance of Long Short-Term Memory (LSTM), bidirectional LSTM (bi-LSTM), Gated Recurrent Unit (GRU), logistic regression (LR), Support Vector Machine with Stochastic Gradient Descent (SVM-SGD), and Support Vector Machine with Support Vector Classification (SVM-SVC) in accurately classifying ECG signals. To enhance the classification accuracy, we explore ensemble methodologies, including the averaging, stacking, and bagging. The evaluation primarily focuses on the comparison of these methodologies based on their respective accuracy values. Additionally, we address the challenges posed by the imbalanced nature of ECG data by analyzing precision, recall, F1-score, support values, and employing the confusion matrix to ascertain the performance of the selected methods. The proposed method was evaluated on the MIT-BIH Arrhythmia Database, demonstrating its efficacy in accurately classifying five different arrhythmias. The results provide valuable insights into the efficacy of these machine learning approaches for ECG analysis, highlighting the significance of ensembled methods in improving classification performance in the presence of imbalanced datasets.