An Automation Detection of Arrhythmia Using DWT-AR Features and Machine Learning

An irregular heartbeat is usually diagnosed using an electrocardiogram, the gold standard for making such a diagnosis (ECG). Researchers in machine learning have proved the volume and diversity of datasets utilized for method development to have a higher effect than the learning algorithm and techniques. In this paper, a new approach is designed using DWT-AR feature selection and classification by the three machine learning techniques. Testing on the MIT-BIH Arrhythmia Database proves the robustness of the proposed algorithm. A hyper-tuning approach is a powerful tool for determining the hyperparameter in ML algorithms. Classifying ECG signals using SVM, ELM, and RVM reveals that RVM provides superior results across all parameters. The suggested program employs a morphological filter during preprocessing to prepare the data better for the DWT-AR modeling used in the ECG beat categorization. Compared to competing methods, it is shown to have superior precision, recall, detection rate, and false alarm to positive ratio (tnratio).