Cardiac arrhythmia classification using hierarchical classification model

Abstract

The application of machine learning techniques in medicine and biomedicine has shown a rising trend and corresponding promising results. Several machine learning techniques, such as artificial neural networks, redial basis function networks and support vector machines, are successfully applied to the classification of different types of heart beat arrhythmia. This paper explores the use of a hierarchical model for the classification of cardiac arrhythmia. Furthermore, it investigates the performance of four machine learning techniques for heart beat arrhythmia classification. The benchmark MIT ECG arrhythmia database is used to evaluate the different models. The results indicate that a TreeBoost based supervised model generally achieves the best performance result. A decision tree forest has comparable results to that of TreeBoost and has slightly higher performance, compared to SVM. However, MLP has the lowest performance result. The results also show that the hierarchical model slightly outperforms the conventional one-stage model in terms of accuracy, sensitivity, and specificity.